Sim vs Microsoft Copilot Studio
Sim is the open-source AI workspace where teams build, deploy, and manage AI agents visually, conversationally, or with code. Here is how Sim compares to Microsoft Copilot Studio on platform architecture, AI capabilities, integrations, pricing, security, and support. Every fact below is sourced and dated.
Sim is an open-source AI workspace for building, deploying, and managing AI agents. This page compares Sim to Microsoft Copilot Studio across platform architecture, AI capabilities, integrations, pricing, security and compliance, observability, and support, using sourced, dated facts for buyers evaluating both platforms.
What is Sim?
Sim is the open-source AI workspace where teams build, deploy, and manage AI agents, connecting 1,000+ integrations and every major LLM to automate real work visually, conversationally, or with code.
What is Microsoft Copilot Studio?
Microsoft Copilot Studio is a low-code Microsoft tool for building, testing, and publishing conversational and autonomous AI agents with topics or LLM-driven generative orchestration, connectors, agent flows, and Dataverse-grounded knowledge.
Sim vs Microsoft Copilot Studio: feature-by-feature comparison
CompareSim vs Microsoft Copilot Studio
Sim
Microsoft Copilot Studio
Platform
Builder type
Visual canvas, chat, or codeVisual drag-and-drop canvas, natural-language (Chat), or code (API/SDK)
Topics or generative orchestration, plus deterministic agent flowsLow-code conversational/agent builder with a topics-based (trigger phrase) authoring mode, an LLM-driven generative orchestration mode, natural-language agent creation, and separate agent flows for deterministic step sequences. Classic authoring uses topics triggered by phrases. Generative orchestration lets the agent dynamically select topics, tools, and knowledge based on a description of each. Agent flows, built on the Power Automate flow engine, handle deterministic, step-by-step logic invoked as agent tools.
Learning curve
Low, plus natural-language Chat for non-technical usersLow for visual building; natural-language Chat surface for non-technical builders. Chat lets users describe a workflow in plain language and have Sim build it.
Easy for a single simple agent, steep for orchestration and ALM at scaleLow for a single-topic agent answering from one knowledge source; steep for generative orchestration instruction design, solution-based ALM, and Dataverse security modeling at production scale. Basic agent creation is accessible to non-developer makers via natural language and templates, but generative-orchestration instructions and solution/environment/pipeline structuring both require deliberate design work, per dedicated Microsoft guidance on each.
Self-hosting
No: Copilot Studio has no self-hosted deployment of its authoring or orchestration/runtime engine; it runs only as a Microsoft-operated cloud service (commercial or government cloud). Copilot Studio is an Online Service across Commercial, GCC, GCC High, and DoD environments, all Microsoft-operated. No on-premises or customer-hosted runtime exists.
Deployment options
Cloud-hosted or self-hosted, no mid-tier VPC optionCloud-hosted (managed, multi-tenant SaaS) or self-hosted (Docker/Kubernetes). No documented managed single-tenant/VPC hosting tier in between. The Enterprise plan's only hosting-related row in the pricing comparison table is a boolean "Self Hosting" flag; there is no dedicated-instance/VPC offering.
Commercial cloud plus GCC/GCC High/DoD government cloudsCommercial multi-tenant cloud, plus Office 365 GCC, GCC High, and DoD sovereign/government cloud environments. Microsoft's SOC 2 compliance documentation lists Copilot Studio among the Power Platform services in scope for Commercial and GCC environments.
Templates
Yes: an Agent Library of ready-to-use, pre-built agent templates (e.g. Employee Self-Service, Prompt Coach, IT Helpdesk, Financial Insights) with preconfigured instructions, actions, topics, and starter knowledge, deployable into an environment and then customized. Templates are distributed as a visual, guided-deployment catalog on Microsoft Marketplace and as raw solution files on GitHub.
License
Apache 2.0Apache License 2.0
Proprietary commercial SaaSProprietary, commercial SaaS, not open source. Copilot Studio is a licensed Microsoft commercial cloud product billed via Copilot Credits (prepaid packs, pay-as-you-go, or annual commitment); no open-source licensing model exists for the product.
Environment promotion
Yes: fork a whole workspace into a dev/qa/prod-style child, diff it, and promote or roll back changes in either direction. Credential and env-var remapping is required before every promote, so secrets are never silently copied across environments. Gated to Enterprise plan on hosted Sim, or a FORKING_ENABLED flag on self-hosted deployments.
Yes: agents built inside a Dataverse solution promote dev to test to production via exported/imported managed and unmanaged solutions, Power Platform Pipelines for automated promotion, and environment variables for per-environment config, the same ALM model Power Automate uses. Solutions are the transport container across environments, with pipelines or Azure DevOps/GitHub Actions automating the import chain and Git integration available for source control.
Version control
Deployment rollback plus Copilot edit diff/revertDeployed-version history with rollback for every workflow; server-persisted checkpoint/revert and visual diff (accept/reject) specifically for Copilot AI edits. Manual drag-and-drop undo/redo is client-side/localStorage only (capped at 100 ops, 5 stacks), not server-synced across devices. Deployment history does not include an arbitrary version-to-version diff tool, and knowledge base documents have no version history.
Git-backed solution promotion, no visual diff view foundSolution-based promotion with Git integration and pipelines for solution-aware agents; no visual diff/compare view between two agent versions, and this ALM tooling does not apply to agents built outside a solution. Git integration lets a solution connect to a repository for version control and collaboration, and pipelines automate deployment between environments, but there is no dedicated side-by-side version-diff UI for an individual agent.
Realtime collaboration
No: Copilot Studio agents can be shared with co-owners and accessed by multiple makers, but there is no live, concurrent multi-user editing of the same agent with visible cursors and synced changes. Microsoft's live coauthoring feature (visible cursors, simultaneous editing) exists for Power Apps Studio canvas apps, a separate product, not for the Copilot Studio agent designer.. Sharing an agent with other users grants asynchronous edit access, not simultaneous live co-editing. Cursor-level real-time collaboration does not exist in the Copilot Studio designer itself.
Native file storage
Yes: a native Files area with folder hierarchy, link-based sharing (public, password, email OTP, or SSO auth), and a workspace-level Recently Deleted view covering workflows, tables, knowledge bases, files, and folders. Admins can restrict which share-auth modes (public/password/email/SSO) a permission group is allowed to use.
No: Copilot Studio has no file storage system of its own (no folder hierarchy, link-based sharing, or recycle bin built into the product). Files used as knowledge sources or by connectors are stored in and managed through external services such as SharePoint, OneDrive, or Dataverse.. Knowledge sources reference SharePoint sites/document libraries or uploaded files stored in Dataverse. There is no first-party, user-facing file manager inside Copilot Studio itself.
Sub-workflows (composition)
Yes: agent flows built in the flow designer include a built-in child-flow action, alongside loop and branch control structures, that calls another published flow as a step, waits for it to finish, and passes/receives data, distinct from an agent merely calling a flow as a tool at the conversation level. The agent flow designer lists control structures for looping and branching, data operations, date and time functions, and child flows as built-in action categories, the same child-flow composition model Power Automate uses for its underlying flow engine.
Pricing
Pricing model
Credit-based billing, BYOK exempt from capsCredit-based usage billing (Stripe), with bring-your-own-key exemption from metered caps
Consumption-based Copilot Credits, several metered feature ratesConsumption-based Copilot Credits, purchasable as prepaid capacity packs, pay-as-you-go against an Azure subscription, or a discounted annual commitment, billed across separately metered feature rates (classic/generative answers, agent actions, tenant graph grounding, agent flow actions, three tiers of AI tools, three tiers of voice). Microsoft 365 Copilot licensed users get included usage at no additional Copilot Credit charge for employee-facing (business-to-employee) scenarios, up to fair-usage limits.
Entry paid plan
Pro plan at $25/user/monthPro: $25 per user/month
$200/month for 25,000 credits, or $0.01/credit pay-as-you-goCopilot Studio Pre-Purchase Plan: $200/month for a 25,000 Copilot Credit capacity pack (roughly 20% cheaper than pay-as-you-go), or pay-as-you-go at $0.01 per Copilot Credit with no upfront commitment. Both require an active Azure subscription for billing; a $200 free Azure credit is available for new accounts to trial the pay-as-you-go meter.
Free tier
No permanent free tier for building/running agents at production scale; a $200 Azure account credit and a Copilot Studio trial are available for new users, and Microsoft 365 Copilot licensed users get included agent usage for internal, employee-facing scenarios. The $200 Azure credit is a one-time trial allowance against the pay-as-you-go Copilot Credit meter, not an ongoing free plan. Microsoft 365 Copilot's included usage only applies to authenticated, licensed-user interactions.
Bring your own key
Yes, for the underlying model rather than a raw provider API key: a maker can bring a specific Azure AI Foundry model deployment (via its endpoint URI, deployment name, and API key) into a prompt as a first-class Bring Your Own Model option, billed separately from the standard Copilot Credit meters. Currently available for individual prompts/tools. Microsoft's release plan lists using a brought-in model as the agent's primary response model, not just in prompts, as a future preview capability.
Security & compliance
SOC 2
Yes: Copilot Studio is audited SOC-compliant, with audit reports available from the Microsoft Service Trust Portal. Copilot Studio's admin-certification documentation confirms SOC compliance without specifying SOC 1 vs SOC 2 vs report Type on that page. The underlying Power Platform SOC 2 Type 2 attestation separately covers Commercial and GCC environments.
Data residency
Full control via self-hosting; Cloud region toggle is global, not per-customerFull data control via self-hosting (Docker/Kubernetes); data never leaves customer infrastructure when self-hosted. On Sim Cloud, async job execution has an internal US/EU region toggle, but it is deployment-wide, not a customer-selectable per-workspace residency option
Yes: organizations can create agents in a specific environment/region so agent data resides within that geography, with Microsoft replicating only within the same geographic area for durability. Covered in Copilot Studio's dedicated geographic data residency documentation, distinct from the general Power Platform environment-region setting.
Role-based access control
Yes: Dataverse-backed security roles (System Administrator, Environment Maker, Bot Contributor, Bot Transcript Viewer, Analytics Viewer, and others) scope who can author, publish, view transcripts for, or view analytics on an agent, assignable directly or via Microsoft Entra ID group teams. Controlling agent creation requires layering tenant-level licensing/Author-group access with environment-level security roles, since neither alone is sufficient. Bot Transcript Viewer specifically scopes access to conversation transcripts.
Audit logging
Yes: Copilot Studio provides dedicated audit logging covering admin, maker, and user activity across agents, viewable in the Microsoft Purview compliance portal. Covered in a Copilot Studio-specific audit-logging article, distinct from the general Microsoft 365 unified audit log Power Automate relies on for the same purpose.
Additional compliance
SOC2SOC2. Self-hosting is the primary lever Sim offers for data-residency-sensitive compliance needs beyond SOC2, rather than additional certifications.
HIPAA, HITRUST, FedRAMP, multiple ISO standards, PCI DSS, CSA STAR, and moreHIPAA (Business Associate Agreement), HITRUST CSF, FedRAMP, multiple ISO standards (9001, 20000-1, 22301, 27001, 27017, 27018, 27701), PCI DSS, CSA STAR, UK G-Cloud, Singapore MTCS Level 3, Korea K-ISMS, and Spain ENS, each with an audit report on the Microsoft Service Trust Portal. This is the full list from Copilot Studio's admin-certification documentation. Each certification links to a corresponding audit report or certificate.
Model & tool governance
Yes: enterprise "permission groups" let an admin allow-list/deny-list specific LLM providers and models, and separately deny specific tools/integrations (or disable all MCP or custom tools) per group, layered on top of workspace admin/write/read roles. This does not control whether an LLM provider retains prompts. Sim offers no "zero data retention" mode or governed AI gateway. A separate, Enterprise-gated feature lets orgs set a log-retention window and redact PII, but that only controls how long Sim itself keeps execution logs.
Yes: admins can enable or restrict which non-default LLMs (e.g. Anthropic Claude) are available tenant-wide, and Data Loss Prevention policies classify connectors, and therefore the tools built from them, into Business/Non-Business/Blocked groups at the tenant or environment level. Model access control is a distinct admin toggle from DLP connector classification. Together they cover both which model an agent may use and which connector-backed tools it may call.
Credential governance
Yes: shared credentials (connected accounts, service accounts, workspace secrets) are their own nested permission level (Member/Admin) below organization and workspace roles, and enterprise permission groups can further allow-list specific integrations and restrict which file-share auth modes (public/password/email/SSO) a group may use. A user's personal environment variables/secrets are never shared or inherited by anyone, including org owners/admins.
No: Data Loss Prevention policies and Advanced Connector Policies govern access at the connector/action/endpoint level (e.g. blocking the SharePoint connector tenant-wide), not at the level of restricting which specific stored credential or connection a given role may use for the same connector. This is the same Power Platform DLP model Power Automate uses, since Copilot Studio agent flows and connectors sit on the same underlying platform.
Single sign-on (SSO)
Yes: Copilot Studio and the wider Power Platform are built on Microsoft Entra ID, which natively supports SAML/OIDC single sign-on plus automatic user/app provisioning (SAML just-in-time and SCIM-based), so signing in via Entra ID grants org-level access without manual per-user account setup. SSO and provisioning are inherited from the Microsoft 365/Entra ID tenant rather than a Copilot Studio-specific setting, consistent with the rest of Microsoft's enterprise stack.
Vetted first-party integrations
Yes: every one of Sim's 302 blocks is first-party authored and code-reviewed through the standard pull-request process in the main Sim repository; there is no public marketplace where an arbitrary third party can publish and have other users install executable tool code without going through Sim's own review. Custom code steps run inside Sim's own isolated-vm sandbox rather than as an installable third-party skill package, so the supply-chain trust boundary is Sim's codebase review, not an open registry.
Certified catalog is vetted; uncertified custom connectors can bypass DLPPartial: the primary connector catalog (1,000+ connectors) is Microsoft-certified, requiring code/security review, functional verification, and malware scanning before publication, including connectors submitted by third-party (independent) publishers. But any maker can also build an uncertified custom connector against an arbitrary API for their own tenant, and security researchers at Zenity have shown that custom connectors can bypass Power Platform Data Loss Prevention policies to reach connectors an admin has explicitly blocked. Microsoft's connector certification process scans all submitted connector code for malware or viruses and runs a functional/security review before a connector joins the shared, browsable catalog. Custom connectors skip that review entirely unless separately submitted for certification, and Zenity's published research demonstrates a custom connector reaching a DLP-blocked connector, a documented real-world weakness in the uncertified path.
PII redaction
Yes, but as a block rather than in-line redaction: Microsoft Purview inline Data Loss Prevention for Copilot Studio agents (public preview) scans prompts sent to an agent in real time for Sensitive Information Types (SSNs, credit card numbers, custom types), and blocks the prompt from being processed, with no AI response generated, if one is detected before the agent is invoked.. This stops sensitive content from reaching the agent at all rather than redacting it in-line and continuing. It is configured as a Purview DLP policy targeting the Copilot Studio location, separate from the product itself.
Custom data retention
28-day default transcript window; longer retention needs a manual exportPartial: session/transcript detail in the Analytics area defaults to a 28-day window (conversation transcripts downloadable for 29 days), extendable only via a separate export pipeline (Azure Synapse Link for Dataverse into Azure Data Lake Storage Gen2) rather than a built-in, admin-configurable retention setting for that transcript detail. This differs from Power Automate's flow-run-history retention, which is directly admin-configurable in the Power Platform admin center (28/14/7 days or a custom Dataverse field edit). Copilot Studio's session-transcript detail requires the export workaround instead.
White-labeling
No: agent-level branding is limited to a custom name, icon, and accent color in Settings, or fully hosting a custom canvas web app for the chat window (via CSS/JavaScript or the Bot Framework SDK). There is no option to remove Microsoft/Copilot Studio branding from the authoring canvas or the default published chat UI itself.. A custom canvas gives full programmatic control over the embedded chat experience, but that is a developer-built replacement UI, not a native white-labeling setting inside the product.
AI capabilities
Multi-LLM support
21 providers incl. OpenAI, Anthropic, Google, Bedrock21 provider integrations (OpenAI, Anthropic, Google/Gemini, Azure OpenAI, Azure Anthropic, Groq, Cerebras, Mistral, xAI, Bedrock, Vertex, Ollama, OpenRouter, and more). apps/sim/providers/models.ts defines 21 provider entries; openrouter/litellm/vllm/ollama resolve models dynamically at runtime rather than from a hardcoded model list.
Yes: agents can use OpenAI GPT models, Anthropic Claude Sonnet 4 and Opus 4.1, any model in the Azure AI Model Catalog, or a bring-your-own Azure AI Foundry model deployment for individual prompts. Admins enable or restrict non-default models tenant-wide in the Microsoft 365 Admin Center. If a selected alternate model is disabled or unavailable, agents fall back automatically to the default OpenAI model. Google Gemini is not supported.
Agent reasoning blocks
Yes: generative orchestration is an LLM-driven planning layer that selects among an agent's topics, tools, knowledge sources, and child agents at runtime, and an optional deep reasoning model (GPT-5.5 Reasoning) can be enabled for tasks requiring step-by-step logical analysis, distinct from a fixed decision-tree topic flow. Deep reasoning is regionally limited to the United States and the EU (excluding the UK) and trades response speed for accuracy on complex tasks. The agent decides when to apply it, or a maker can force it via a "reason" keyword in instructions.
Natural-language building
Yes: makers can describe an agent in plain language and Copilot Studio's AI-based authoring drafts topics, instructions, and structure from that description, distinct from manually assembling a topic tree. This is the product's core AI-based agent authoring capability, letting non-developers describe the desired behavior instead of hand-building every topic and trigger phrase.
Knowledge base / RAG
Yes: agents can be grounded via retrieval-augmented generation over Dataverse tables, SharePoint/Office files, connectors to systems like Salesforce or ServiceNow, and a tenant-wide semantic search index (tenant graph grounding) over Microsoft Graph data including connector-synced external content. Tenant graph grounding is a separately billed, optional per-agent feature (10 Copilot Credits per message) layered on top of ordinary knowledge-source retrieval, for higher-quality, up-to-date grounding.
MCP support
Yes: agents connect to external MCP servers as tools directly from the Tools page (Add Tool > New Tool > MCP, specifying just a URL), and tools/resources the server publishes are automatically added as actions that inherit their name, description, inputs, and outputs, staying in sync as the server changes. MCP servers are wired in via connector infrastructure under the hood, so they inherit enterprise controls like Data Loss Prevention policies, virtual network integration, and multiple authentication methods.
Evaluation & guardrails
LLM-judge Evaluator plus Guardrails validation blockEvaluator block (LLM-judge scoring against user-defined named metrics) and Guardrails block (JSON validity, regex, RAG/hallucination scoring, PII detection/masking). These are per-call scoring/validation primitives, not a batch golden-dataset eval-suite runner or A/B prompt-testing harness.
Yes: a dedicated test-set evaluation feature runs up to 100 test cases (authored manually, imported from a spreadsheet, or AI-generated from the agent's design/knowledge/topics) against an agent, alongside mandatory content moderation that screens both user input and agent responses for harmful content, jailbreaking, prompt injection, and prompt exfiltration. Content is evaluated twice per turn (at input and again before the response is returned); a detected violation blocks the response and shows an error rather than letting it through.
Human-in-the-loop
Yes: a dedicated Request for Information action in agent flows pauses an automated run to collect structured input from a designated human reviewer via email, and a separate hand-off feature lets a conversational agent transfer an in-progress chat, with full history and variables, to a live human agent in a connected engagement hub. Request for Information targets agent-flow-style automation pausing on a decision point, and hand-off targets live conversational escalation to a human, covering both the automation-approval and conversational-escalation senses of human-in-the-loop.
Generative media
Captioning and text gen only; image/audio generation needs an external connectorPartial: AI Builder, the shared Power Platform model catalog Copilot Studio prompts draw from, ships an image-description (captioning) model and GPT-based text generation, but no native image-generation, video-generation, or text-to-speech/speech-to-text model. Generating images or audio requires an external connector, such as Azure OpenAI DALL-E or Azure AI Speech.. Copilot Studio prompts and agent flows both draw from the same AI Builder catalog Power Automate uses, so the gap in native generative-media models is shared across both products.
Dynamic tool use
No: an Agent block calls tools the workflow author explicitly added to it at build time, rather than browsing and picking from a broader pool (e.g. an entire MCP server catalog) at inference time. Runtime MCP "discovery" exists to resolve/refresh the schema of an already-configured tool. The model does not browse or choose from the server's full tool list.
Yes: with generative orchestration turned on, an agent's LLM-driven planner dynamically selects which topics, tools, and knowledge sources to invoke at runtime from the full set attached to the agent, rather than following only the specific tool-call path a maker pre-wired into a fixed topic. This is what generative orchestration adds over classic topic-based authoring, where a maker had to hand-author which tool a given trigger phrase path calls.
Automatic model fallback
No: a failed or rate-limited LLM call is retried using Sim's own hosted API keys for the same model, rather than automatically switching to a different model or provider. A "fallback" comment in the provider layer refers to rotating among Sim's own hosted API keys for the same model, not switching models.
Yes: Copilot Studio automatically falls back to the default OpenAI GPT-4o model when a selected alternate model, such as Anthropic Claude, is disabled or unavailable for the tenant. This is a multi-model fallback behavior, not a broader multi-provider retry-on-rate-limit policy beyond that single fallback path.
Agent skills
Yes: named, reusable "Agent Skills" (built on the open Agent Skills / SKILL.md format) that agents load on demand via progressive disclosure, editable in-app or imported from a SKILL.md file or GitHub URL. Only the skill name and description sit in the agent's system prompt (~50-100 tokens each); the full instructions load into context only when the agent calls load_skill.
Yes: Skills are named capabilities (name, description, Markdown instructions, optional supporting files) defined once, created in Copilot Studio or a text editor, added to multiple agents, and exported as a Markdown file or ZIP package for sharing, distinct from a single agent's own instructions. This is a preview feature in the new Copilot Studio agent-building experience. The standard SKILL.md format with YAML front matter makes skills portable between agents and organizations.
Native chat deployment
Yes: an agent can be published directly to a live website chat widget, Microsoft Teams, Microsoft 365 Copilot Chat, SharePoint, Power Pages, mobile, and further channels via Azure Bot Service (Slack, Telegram, Twilio SMS, and more), not only a form/API/webhook. Publishing pushes the current agent version out to every connected channel at once. A channel is "the integration point where an end-user can interact with a Copilot Studio agent."
Parallel execution
Yes: a native Parallel block fans a run out into concurrent branches (fixed count or one per list item) and joins their results back into the workflow automatically. Contained blocks run concurrently instead of sequentially, either a fixed number of times or once per item in a list/collection, and each branch's output aggregates for downstream blocks.
Yes: agent flows, built on the Power Automate flow engine, support native parallel branches so multiple actions run concurrently and the flow proceeds only once every branch completes; loop actions separately support a parallel (concurrency) mode for independent iterations. Parallel branches and concurrent loop iterations add complexity and can affect output quality when combining concurrent results, and flows must still return within a documented action-count/time budget.
Agent2Agent (A2A) protocol
No native support: Copilot Studio does not ship a first-party Agent2Agent (A2A) implementation today. Third-party custom connectors can wrap an external A2A agent's JSON-RPC/HTTP+JSON endpoints, and Microsoft has stated native A2A support is planned for Copilot Studio and Azure AI Foundry, but no built-in Agent Card discovery or native peer-to-peer A2A calling ships today.. Available A2A connectors today are community-built, translating Power Platform requests into A2A protocol calls, not a first-party Copilot Studio feature.
Loop / iteration block
Yes: agent flows include a built-in Loop action supporting both For each (iterating a known collection) and Do until (repeating until a condition becomes true) patterns, running sequentially by default with an optional parallel/concurrency mode (a configurable degree-of-parallelism setting) for iterations that are fully independent. Operations accumulating a value or depending on order must stay Sequential, since parallel mode risks race conditions on shared variables and API throttling. Do until loops also require a maximum iteration count and timeout to prevent runaway execution.
Integrations
Integrations
302 blocks, ~3,900 tool actions302 first-party blocks, ~3,900 underlying tool actions. Sim's landing page cites "1,000+ integrations," a broader figure counting individual API actions rather than top-level blocks. Both numbers describe the same integration surface.
1,000+ connectors from the shared Power Platform catalog1,000+ pre-built connectors. Copilot Studio shares the same underlying connector catalog Power Automate uses, whose product page cites 1,400+ certified connectors as a broader Power Platform-wide figure.
Trigger types
Webhook, cron, chat, REST API, triggers for 61 appsWebhook, schedule/cron, chat, REST API, and event-based triggers for 61 apps (Slack, Gmail, GitHub, Stripe, etc.)
Trigger phrases, autonomous event triggers, connector events, schedulesConversational trigger phrases (topics), autonomous event-based triggers that wait for a specific event to fire the agent without a user prompting it, connector-event triggers, and schedule-based triggers for agent flows. Autonomous triggers let an agent proactively respond to events, such as a new record or an incoming email, rather than only reacting to a live conversation.
Custom code steps
No generic inline script step in the conversational designer; custom logic is reached via agent flows composed of Power Automate flow actions, or custom connectors wrapping an Azure Function or REST/SOAP API. Agent flows are built with the same low-code flow-action model as Power Automate, not an arbitrary code editor. A custom connector can front an Azure Function to run bespoke code as a callable action.
API publishing
Yes: a published agent is reachable over the Azure Bot Service Direct Line API/channel, giving external code a callable HTTP interface into the conversation, alongside the standard chat channels. Direct Line is the same Bot Framework mechanism used to embed a custom canvas or drive an agent from an external application, not a distinct, separately branded API-publishing feature.
SDKs & extensibility
Bot Framework SDK, custom canvas, pac CLI, open-source connector repoMicrosoft Bot Framework SDK and custom-canvas hosting for programmatic chat-UI control, the Power Platform CLI (pac CLI) for solution/pipeline scripting, and the shared, open-source microsoft/PowerPlatformConnectors GitHub repo for community connector submissions. No separate, independently monetized third-party marketplace exists beyond the shared certified-connector catalog. Custom canvas hosting lets a developer take full programmatic control of the chat surface via Bot Framework.
Publish as MCP server
No: Copilot Studio primarily consumes MCP servers, adding an existing server's tools/resources to an agent, the reverse direction. No feature lets a maker publish a Copilot Studio agent itself as a callable MCP server for external AI tools to invoke.. Exposing an MCP server for Copilot Studio to reach means hosting a custom MCP server separately (e.g. via Azure Container Apps or VS Code dev tunnels) and connecting Copilot Studio to it as a client, not publishing the agent as a server.
Observability & durability
Tracing & observability
Yes: execution logs include a per-block/per-span trace view (duration, cost, token counts, and latency stats like TTFT/TPS) with expandable nested iteration groups, plus a "View Snapshot" frozen copy of the workflow structure and block states at run time for debugging. This trace view is built directly into Sim rather than a raw export browsable in an external tool like Jaeger, and does not expose aggregate latency-percentile charts (p50/p95/p99). The run snapshot serves as a log-detail/debugging artifact rather than a resumable mid-run checkpoint.
Yes: per-session conversation transcripts show which topic fired, which knowledge sources were consulted, which tools were called, which child agents or MCP servers were invoked, what the orchestration plan was, and how long each step took, alongside an Analytics dashboard with conversation volume, engagement, satisfaction, and response-quality metrics over time. Telemetry can additionally be sent to Azure Monitor Application Insights, where a dedicated Copilot Studio dashboard workbook surfaces total conversations, latency, exceptions, tool usage, and topic analytics in one view for deeper analysis.
Durability & retries
Tool-call retries (up to 10x); single-attempt job orchestrationIndividual tool/API calls have configurable exponential-backoff retry (up to 10 attempts). The background job-orchestration layer itself retries only once by design. Durability instead comes from consecutive-failure tracking on schedules and the human-in-the-loop snapshot pause/resume mechanism. Sim does not offer guaranteed-once-only block execution, a failed-run holding queue for manual recovery, or a "replay a past execution with its original inputs" feature. The per-execution debugging snapshot serves as a log-detail artifact rather than a resumable mid-run checkpoint.
Yes for agent-flow steps: because agent flows are built on the Power Automate flow engine, individual flow actions can use configurable retry policies with backoff, the same durability mechanism Power Automate exposes. A documented troubleshooting pattern for tool-call timeouts explicitly recommends adding retry with backoff. This durability applies to the deterministic agent-flow layer. No separate, distinct replay/checkpoint mechanism exists for the conversational/generative-orchestration layer itself.
Failure alerting
Alerting requires configuring Application Insights; no native failure-email feature foundPartial: proactive failure alerting is reachable through Azure Monitor Application Insights alerts on exceptions/latency once telemetry is wired up, but there is no Copilot Studio-native, automatic per-run failure-email or weekly-digest feature comparable to Power Automate's flow-failure notifications.. Copilot Studio's Analytics area is dashboard/lookup-based, so a maker must open it to see failures, while pushing a notification depends on separately configuring Application Insights alert rules.
Data drains
Yes: Enterprise orgs can continuously export workflow logs, job logs, or audit logs on a schedule to a customer-owned S3 bucket, GCS bucket, Azure Blob container, BigQuery table, Snowflake table, Datadog logs intake, or an HTTPS webhook. Each drain exports exactly one data source; multiple drains are created to export multiple sources. Viewing drain config/run history is restricted to org owners/admins.
Yes: telemetry can be continuously streamed to Azure Monitor Application Insights, and conversation transcripts/custom analytics data stored in Dataverse can be continuously exported via Azure Synapse Link for Dataverse into Azure Data Lake Storage Gen2 in Common Data Model format. This is the same Azure-native export pattern used across Power Automate and Power Platform more broadly (Microsoft Sentinel for audit/activity logs, Application Insights for telemetry), not a generic user-configurable webhook/S3/BigQuery drain built directly into Copilot Studio.
Async execution
Yes: a workflow can be triggered in fire-and-forget async mode, returning HTTP 202 with a job ID immediately, then polled via a dedicated jobs endpoint through queued/processing/completed/failed states. Async jobs are tracked via polling the job endpoint rather than a completion webhook/callback option.
Yes: agent flows inherit the Power Automate flow engine's asynchronous response pattern (an HTTP 202 plus a Location header the caller polls), letting a long-running flow action continue beyond a synchronous request's time limit. Agent flow actions invoked by an agent are still bound by a 100-second action limit within a conversational turn, distinct from the longer-running asynchronous pattern available to a flow triggered independently of a live conversation.
Execution limits
5-50 min sync timeout, 90 min async, 15-300 concurrentPlan-gated: synchronous API calls time out at 5 minutes on the free plan and 50 minutes on paid plans, async calls at 90 minutes on every plan, with 15 to 300 concurrent executions per billing entity depending on plan. These limits are not published in docs; request bodies are separately capped at 10 MB.
~100-120s per-turn action limit, 100-action/64KB express-mode capMicrosoft publishes concrete limits: agent flow actions invoked by an agent must respond within roughly 100-120 seconds per conversational turn; express-mode flow runs are capped at 100 actions and a 64 KB message size per action; and generative AI/Copilot Credit throttling applies per Dataverse environment when consumption exceeds capacity.. These limits are specific to the agent-flow/tool-calling layer inside a live conversation. The underlying flow engine also carries broader Power Automate execution limits (30-day max run, etc.) when a flow runs independently of an agent turn.
Partial-failure handling
Yes for agent-flow steps: because agent flows run on the Power Automate flow engine, an individual failing action can be routed to an error-handling path via the same per-action "Configure run after" (has failed / is skipped / has timed out) setting, letting the rest of the flow continue rather than the whole run always halting. This inherits directly from the Power Automate flow-action model agent flows are built on. No separate, distinct partial-failure mechanism exists for the conversational/topic layer.
Support
Support channels
Community support plus Enterprise 'Dedicated Support'Community (open source, GitHub) plus an unquantified "Dedicated Support" flag on the Enterprise plan. Enterprise and pricing pages do not include CSM, onboarding/enablement, or professional-services details beyond a plan-comparison-table "Dedicated Support" flag.
Docs, community/Q&A forums, and Admin Center for business-critical issuesDocumentation via Microsoft Learn, the Power Platform/Power Users community forum and Microsoft Q&A for general questions, and the Microsoft 365 Admin Center for business-critical, SLA-based support requests. Microsoft's guidance distinguishes routine community/Q&A support from business-critical issues, which go through the Microsoft 365 Admin Center support flow rather than the public forums.
SLA
No product-specific SLA found; general Online Services SLA appliesNot publicly documented as a Copilot Studio-specific, financially backed SLA; general Microsoft Online Services SLA terms (covering Azure, Dynamics 365, Office 365) apply, with a widely cited 99.9% uptime commitment for other core Microsoft 365 services. Reporting on Copilot outages has noted enterprise customers lack the same financially backed SLA protection for Copilot that exists for core services like Exchange Online or file storage.
Community
100,000+ buildersOver 100,000 builders use Sim
Large, shared Power Platform community forum plus Microsoft Q&ALarge: an active, Microsoft-hosted Power Platform/Power Users community forum with structured Q&A on building, publishing, and troubleshooting Copilot Studio agents, plus Microsoft Q&A for developer-specific questions. The same community forum infrastructure serves the broader Power Platform (Power Apps, Power Automate, Copilot Studio), not a dedicated, separately branded Copilot Studio-only forum.
Academy / training
Yes: Microsoft Learn provides structured, self-paced training paths, hands-on labs (Microsoft Copilot Studio Labs, Agent Academy), and official certifications spanning Copilot Studio and the broader Power Platform, including Microsoft Certified: Power Platform Fundamentals (PL-900). Beyond formal certification paths, Microsoft publishes dedicated, scenario-based hands-on labs (mcs-labs, agent-academy) specifically for Copilot Studio, going beyond ad hoc docs or blog content.
Sim standout features
AI Copilot / Chat agent-building surface
Chat and in-editor Copilot suggest and build workflow changes directly.
A natural-language surface (Chat) and in-editor Copilot that can explain, suggest, and build workflow changes directly, backed by a dedicated copilot module with its own tool registry.Hybrid semantic + keyword knowledge base
Combines vector and full-text search with configurable chunking across 11 file formats.
Built-in RAG with pgvector embeddings and a generated tsvector column for combined vector + full-text search, plus a token-based chunker with configurable chunk size/overlap and 11 supported file formats (csv, doc, docx, html, json, md, pdf, pptx, txt, xlsx, yaml).Native MCP client and server
Call external MCP servers as tools, or expose Sim workflows as an MCP server.
A dedicated MCP block lets any workflow call external MCP servers as a tool, and a serve/workflow-servers API surface lets Sim expose its own workflows as MCP servers.Fork a workspace into dev, qa, and prod environments
Fork, diff, and promote environments with mandatory credential remapping.
Fork a whole workspace into a dev/qa/prod-style child environment, preview a diff, and promote changes bidirectionally. Credential and env-var remapping is required on every promote, so secrets never cross environments silently.Human-in-the-loop approvals with durable resume
Pause a run for human approval and resume later via a durable snapshot link.
A dedicated block pauses a run and waits for a human-submitted approval form, backed by persisted execution snapshots so the run can resume later via a link, even after a server restart.Self-hostable under Apache 2.0
Fully open source with Docker Compose and Helm deployment options.
Fully open source (Apache 2.0), with Docker Compose files and a Helm chart for Kubernetes deployment, alongside a managed cloud-hosted option.Documented Microsoft Copilot Studio limitations
Consumption pricing carries overage and cost-forecasting complexity
Costs span many separately metered credit rates, needing a dedicated estimator tool.
Usage is billed in Copilot Credits across many separately metered feature rates (classic answer, generative answer, agent action, tenant graph grounding, agent flow actions per 100, three tiers of AI tools, three tiers of voice), and reasoning models add a second, separate premium-token charge on top of the base rate. Microsoft publishes a dedicated usage estimator tool, since forecasting cost across these dimensions manually is difficult.Full ALM tooling (pipelines, environment variables) is gated to solution-aware agents
Pipelines and environment-variable promotion require building inside a solution.
Version history, environment promotion via pipelines, and environment variables only apply to agents built inside a Dataverse solution. Agents created outside a solution, the common default for individual makers experimenting in Copilot Studio, get none of this ALM tooling.Session transcript detail defaults to a 28-day window
Detailed session transcripts default to 28 days; longer retention needs a manual export.
An agent's per-session transcripts, showing which topic fired, tools called, and knowledge consulted, are available in the Analytics area for the last 28 days by default. Retaining that detail longer requires a separate export pipeline (Azure Synapse Link for Dataverse into Azure Data Lake Storage Gen2), not a built-in retention setting.No self-hosting; a Microsoft-operated cloud service only
No on-premises option; agents run only on Microsoft-operated cloud infrastructure.
Copilot Studio's authoring, orchestration, and runtime are Microsoft-operated multi-tenant (or government-cloud) services. There is no on-premises deployment option for the core agent-building and conversation-orchestration engine, unlike products that offer a self-hosted or air-gapped runtime.No native white-labeling of the core builder or default chat UI
Branding is limited to name/icon/color or a custom-built chat canvas, not full white-label.
Agent-level branding is limited to setting a name, icon, and accent color, or hosting a fully custom canvas web app for the chat window. There is no option to remove Microsoft/Copilot Studio branding from the authoring canvas or the default published chat surface, unlike a dedicated white-label offering.Bottom line
Choose Sim if you want an open-source, self-hostable AI workspace that treats AI agents as first-class citizens: native multi-LLM support, real-time multiplayer editing, environment promotion (dev/qa/prod), human-in-the-loop approvals, and enterprise governance (SSO, credential-level permissions, audit logs) built in rather than bolted on.
Choose Microsoft Copilot Studio if you specifically need reusable, portable Agent Skills: A Skill is a named capability defined once as a SKILL.md file (YAML front matter plus Markdown instructions, optionally bundled with scripts, templates, or reference documents into a ZIP package). Skills are authored in Copilot Studio or a text editor, attached to multiple agents, and exported to share with others, unlike a one-off system prompt tied to a single agent.
Frequently asked questions
Sim is an open-source AI workspace where teams build, deploy, and manage AI agents visually, conversationally, or with code. Microsoft Copilot Studio is a low-code Microsoft tool for building, testing, and publishing conversational and autonomous AI agents with topics or LLM-driven generative orchestration, connectors, agent flows, and Dataverse-grounded knowledge. Teams considering a switch typically weigh licensing (Sim is Apache 2.0 and self-hostable), pricing model, and how AI-native the platform's agent-building experience is.