Microsoft Copilot Studio MCP

# Fix list: Microsoft Copilot Studio MCP

From Anchor Terminal's listing at https://www.anchorterminal.com/tools/copilot-studio-mcp, the October 2026 research run, assessed 17 October 2026. Grade D, 53.1 out of 100.

This is everything the published grade says the listing lacks, the biggest possible gain to the total first. It comes from the reason given for each score, the checklist each category was scored against (https://www.anchorterminal.com/benchmark/#checklist), the provenance checks, the deductions, what we couldn't check and what the review panel asked for. A fix counts at the next check, once it's public.

For a coding agent working on Microsoft Copilot Studio MCP: work through the items below in the product, its docs and its public pages. Each category gives the reason for its score, with the points each checklist item earned, and the checklist itself, so the gap is the items that earned less than their points. Change the product, not the wording, and keep a note of what you changed and where it's published.

## 1. Reliability, 46 out of 100, up to 10.8 more on the total

Why it scored 46: Read with the local-software lines, as for azure-mcp, because the owner runs the stdio bridge, though every tool past sign-in is served by Copilot Studio. The package is one day old. Installs from npm under Microsoft's account with Node.js 22.18 or later stated (20). The repository holds no source and no workflows. The package.json embedded in the bundle names test and coverage scripts with 80 per cent thresholds, but no public CI or test run exists to read (0). The issues page answered 403 to our reader. The repository's only activity is two README pull requests merged on 12 and 13 October 2026, and a second release followed the first within 19 hours, with no note saying why (8 of 25). Semver numbering, but no changelog, release notes or tags (3 of 15). Version 1.0.1 (15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-reliability):

Hosted APIs, MCP servers, models and platforms.

- 20, a public status page with component history (Statuspage, Instatus, BetterStack or the vendor's own).
- 0 to 30, the incident record for the last 90 days on that page. 30 for a clean record or trivial incidents only, 20 for minor incidents only, 10 for one major outage (an hour or more of a core API down, or errors across the board), 0 for several. 5 when there's no history we could read, and the note says so.
- 15, rate limits documented with numbers.
- 15, documented 429 or overload handling (Retry-After, backoff guidance), and idempotency keys or safe-retry guidance where writes are involved.
- 10, an SLA published for any paid tier.
- 10, the surface agents use is generally available, not beta or preview.

Local packages, SDKs, frameworks and stdio MCP servers.

- 20, installs from an official package with supported runtimes stated.
- 25, a public CI and test suite, passing on the default branch.
- 0 to 25, open crash or regression issues relative to activity (25 for few and handled, 0 for many, old and unanswered).
- 15, semver discipline and breaking changes called out in a changelog.
- 15, version 1.0 or later, or declared stable.

Protocols are read from their reference implementations, the public facilitators or servers, spec stability and test vectors.

## 2. Payments & pricing, 20 out of 100, up to 10 more on the total

Why it scored 20: No payment protocol (0). The bridge is free, so read with the rule for the paid service it needs. Copilot Credit packs are public at $200.00 for 25,000 credits a month, but credits per action vary and are published only as ranges and an estimator (10 of 20). An individual can start a Copilot Studio trial with a work account and no card requirement was stated, but the trial article covers the standard harness and trial agents can't be published, and GitHub Copilot harness usage is billed from the first build (10 of 20). A person signs in through a browser, and credits are allocated to environments by an administrator (0).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-payments):

The published rubric, also on the [x402 page](https://www.anchorterminal.com/x402/).

- 40, a machine payment protocol (x402, MPP or L402) on the tool's own endpoints. 10 to 30 when it covers only some endpoints or only goes through a third party, and the note says which.
- 20, per-call or per-unit pricing published without a login. 10 for public plan-only pricing, 0 for "contact sales" or prices behind a login.
- 20, a free tier or trial that doesn't need a card.
- 20, autonomous onboarding, meaning an agent can get access without a person signing up in a browser (keyless use, x402, a programmatic key API).

Payment platforms and agent wallets rarely charge for their own API over a machine protocol, so the first line has steps for them, and the highest one that applies counts. 40 when x402, MPP or L402 runs on all their own endpoints, 30 when it runs on part of their own API, 25 when their merchants can accept one, 20 for running a facilitator, 15 for paying as a buyer, and 0 when the only protocol is their own. Merchant acceptance sits above a facilitator because the platform's own customers can charge agents through it, while a facilitator settles for sellers who wire up the protocol themselves. The counter-argument (a facilitator does more for the protocol as a whole) has a point. Each note says which step applied.

Open-source software you run yourself is scored on its hosted or paid option if it has one. A free, self-hosted package with nothing to buy gets 20, 20 and 20 for the last three lines, and 0 to 40 for the first only if it ships a payment protocol.

## 3. Schema & documentation, 53 out of 100, up to 7.6 more on the total

Why it scored 53: The nine local tools have JSON Schema inputs with `additionalProperties: false`, required fields, regex patterns, a UUID format on plan IDs and `const: true` on confirmations, read from the bundled source. The five service tools are loaded after sign-in and weren't read, and the README says they take `action` and `parameters` (15 of 25). No llms.txt and no Learn page for the bridge. The README is Markdown on npm, and authoring guidance arrives from the service as Skills with a `schema.json` (5 of 10). Local descriptions say when to call each tool and when not to, such as not reconnecting a ready session. The service tools get one line each in the README (15 of 20). Local inputs are tightly typed. The service tools are dispatchers whose `parameters` contract is read at run time through `describe_actions` (8 of 15). The README gives a first prompt and failure guidance, and the source has coded error messages, but no error catalogue is published (7 of 15). Semver, no changelog (3 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-schema):

APIs and MCP servers.

- 25, a machine-readable contract (a public OpenAPI file or similar; for MCP, typed JSON Schema inputs on every tool).
- 10, llms.txt or Markdown docs served for agents.
- 0 to 20, descriptions that say what a tool is for, when to use it and when not to, read from the tool definitions in the source or the API reference.
- 0 to 15, typed inputs with enums, constraints and required fields, and no free-form JSON blobs.
- 0 to 15, examples and documented error responses.
- 15, versioning and a public changelog.

Models are read from the API reference, the OpenAPI file, llms.txt, the structured-output and tool-use docs and the model cards. Frameworks from docs a model can follow, typed interfaces, examples and the API reference.

## 4. Agent ergonomics, 62 out of 100, up to 6.2 more on the total

Why it scored 62: Nine local tools before connection and five service tools after, 14 in all (15). Only setup tools load until an environment is selected, and authoring guidance loads on demand as Skills, which adds back a little. The local definitions run to about 8,500 characters of source and the server instructions to about 7,500, which costs context (2 added back). `copilot_studio_list_environments` takes a name query and full IDs bypass the 1,000-result limit. Paging and size controls on the service tools weren't readable (8 of 20). Errors come back as `isError` results that say whether a write may have completed and whether retrying is safe, with distinct codes for DNS, refused, reset and timed-out connections (16 of 20). Every local tool carries readOnlyHint, destructiveHint and idempotentHint, and connector writes go through a plan ID and `confirm: true`. No idempotency keys, and the service tools' annotations weren't readable (13 of 20). Zero configuration beyond `npx`, with no client ID or tenant to set and most local tools taking no arguments. One Node.js package and no SDK in a second language (8 of 15).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-ergonomics):

- 0 to 25, context cost. For MCP, the number and size of the tool definitions (25 for ten or fewer compact tools, 15 for 11 to 30, 5 for more than 30, plus up to 10 back for toolsets, dynamic loading or read-only subsets). For APIs, whether responses can be sized (field selection, limits, summaries).
- 20, pagination, filtering and output-size controls.
- 20, actionable, documented error responses, codes and messages an agent can recover from.
- 20, idempotency or safe retries, and for MCP the `readOnlyHint` and `destructiveHint` annotations.
- 15, sensible defaults, few required parameters, and official SDKs in at least two languages.

Models are read for tool use, structured output, prompt caching, context length, batch and SDKs. Frameworks for how much code and how many defaults a tool-calling agent with MCP needs.

## 5. Security & auth, 70 out of 100, up to 5.3 more on the total

Why it scored 70: Read with the API and MCP lines, since the bridge runs locally but every call acts on a hosted service. Microsoft Entra delegated OAuth with Power Platform scopes requested per step (environment read, Copilot Studio execute, connector read and write), tokens held only in memory, personal accounts refused and requests pinned to the expected Power Platform hosts. The bridge borrows the VS Code first-party client, so a tenant can't consent to or restrict a dedicated app for it (26 of 30). Connector creation or repair needs a prepared plan and `confirm: true`, and the instructions tell the host to show ring, account, environment and action first. There is no read-only mode, and agent and workflow writes, including workflow tests that run live workflows, go through the service tools with no confirmation step found (11 of 20). `copilot_studio_list_environments` tells the model to treat returned text as data. Server instructions and Skills from the service are passed to the host as guidance, and we found no wider prompt-injection guidance (7 of 15). Copilot Studio maker audit logs are in Microsoft Purview and Sentinel, and `get_consumption` reports credit use per agent or workflow (11 of 15). SECURITY.md routes reports to MSRC, which runs a bounty and publishes CVEs, including two for Copilot Studio this year (CVE-2026-21520, 22 January 2026, CVSS 7.5, and CVE-2026-80098, 3 September 2026, CVSS 9.3). microsoft.com's security.txt expired on 23 September 2026, the npm releases carry no provenance attestation, and Copilot Studio's compliance page wasn't read (15 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-security):

- 0 to 30, the credential model. 30 for OAuth 2.1 with scopes, or scoped and revocable keys with rotation. 20 for plain revocable API keys. 10 for one all-powerful key. 10 off when a secret can travel in a URL query string as a documented option.
- 0 to 20, read-only or least-privilege modes, and confirmation or approval for destructive actions.
- 0 to 15, prompt-injection posture where the tool returns untrusted content (documented mitigations or guidance). A tool that returns no untrusted content gets 10.
- 0 to 15, audit logs or per-call visibility for the operator.
- 0 to 20, a security programme. security.txt or a disclosure policy, a bug bounty, SOC 2 or ISO 27001, advisories handled in public.

Models are read for retention, whether API data trains models (and whether that's off by default), zero-retention options and certifications. Frameworks for telemetry defaults, approval hooks, guardrails and sandboxing.

## 6. Maintenance & community, 55 out of 100, up to 3.9 more on the total

Why it scored 55: 1.0.1 on 16 October 2026 (30). Two releases, both on 16 October, plus a placeholder version (0). SUPPORT.md is Microsoft's unedited template with placeholder text for where to get help, and the issues page was unreadable to us. The two README pull requests were merged by their author the same day (5 of 25). `com.microsoft/copilot-studio` is active in the official MCP registry under a DNS-verified namespace (15). The bundle pins current dependencies (msal-node 7.0.0, MCP SDK 2.0.0, zod 4.6.5) and needs none at install, but no CI or provenance is visible (5 of 10).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-maintenance):

- 0 to 30, time since the last release, or the last published model or API change for a closed service. 30 within 30 days, 20 within 90, 10 within 180, 0 older.
- 20, at least three releases or dated changelog entries in the last 90 days.
- 0 to 25, responsiveness. Issues and pull requests answered on GitHub (the open issues and how recent the replies are). For closed services, a public changelog and a support or community channel that answers, 0 to 15.
- 15, presence in the official MCP registry under a verified namespace (MCP servers), or current official SDKs (APIs and models).
- 10, package health, current dependencies and CI.

Models are read for deprecation notice periods and model churn rather than release counts.

## 7. Transparency & trust, 65 out of 100, up to 3.1 more on the total

Made of editorial 47, provenance 83.

Why it scored 65: The repository says MIT and holds no code. The npm package, which is what runs, has no `license` field and no licence file. The bundle is readable but not published as source under a licence (8 of 30). The README's data-handling section says tokens stay in memory, no diagnostics are captured and only a normalised host-family label is sent to the service, and the bundled source agrees (`credentialsPersisted: false`, diagnostics refused in the public build). The service side falls under the Product Terms and the Data Protection Addendum, which we didn't read today (22 of 30). No deprecation policy for the bridge and no notes on what changed between versions (3 of 20). The host-family label is disclosed with no opt-out. Copilot Studio documents data residency and the outbound telemetry Customer Lockbox doesn't cover, but no subprocessor list was read (14 of 20).

The checklist (https://www.anchorterminal.com/benchmark/#checklist-transparency):

- 0 to 30, source availability and licence clarity. 30 for open source under an OSI licence, 15 for closed with clear terms, 0 for unclear terms.
- 0 to 30, data handling and retention statements that agree with each other (privacy policy, DPA, retention periods, subprocessors).
- 0 to 20, a deprecation policy or notices with dates.
- 0 to 20, telemetry disclosed with an opt-out (local software), or subprocessors and data locations disclosed (hosted).

The other half of Transparency and trust is the provenance score, computed from checked facts (below). The category score is the mean of the two.

Provenance checks not met in full (half of this category, computed from checked facts):

- Privacy policy: read, states 8 of the 8 things a reader expects, and has 1 clause that costs points (8 of 10)
- Changelog: not found (0 of 10)
- security.txt: published but past its Expires date (5 of 10)

## What we couldn't check

What we couldn't read counted as absent. Publishing it on a page a plain HTTP fetch can read (not only in a browser) lets the next check count it.

- unchecked: the five service tools' schemas, descriptions and annotations, which load only after a Microsoft work-account sign-in
- unchecked: GitHub issues, pull requests and stars, because github.com pages and the API answered 403 to our reader. The repository was read through a shallow clone
- unchecked: whether the service behind the bridge is the GitHub Copilot harness experience, labelled production-ready preview, or the generally available standard harness. The README doesn't say
- unchecked: the Microsoft Product Terms, the Data Protection Addendum and Copilot Studio's compliance offerings page
- unchecked: the pay-as-you-go price per Copilot Credit and whether failed or cancelled calls consume credits
- unchecked: whether the Copilot Studio trial asks for a card and covers the tools this bridge reaches
- Two Copilot Studio service CVEs from the last 12 months (CVE-2026-21520 and CVE-2026-80098) were fixed and published by MSRC before this package existed. They are recorded under Security and not deducted. A lander may weigh them differently
- No npm download count yet. api.npmjs.org returned not found for the package on 17 October 2026

## Weaknesses

- First published on 16 October 2026, with two releases and no release notes or changelog
- The repository holds no source, and the npm package carries no licence, so the MIT label covers only an empty repository
- The five service tools are dispatchers taking `action` and `parameters`, and their schemas are readable only after sign-in
- No headless sign-in. Each restart needs a browser sign-in with a work account
- SUPPORT.md is Microsoft's unedited template, so no support channel is named

## What costs an agent a turn today

The notes we give agents before they call it. Each one is a workaround an agent shouldn't need.

- Call `copilot_studio_status` first and reuse a ready session. Don't reconnect or reselect an environment to start a new task
- Read an action's contract with `discover_resources` and `action=describe_actions` before building unfamiliar `parameters`
- Ask the person before testing a workflow. Tests run published, enabled workflows and can have real effects
- Never retry an uncertain write. Inspect the agent, workflow or connection first, as the error text says
- Expect a browser sign-in after every restart. Tell the person a window is opening and never ask for credentials in chat

## What the review panel asked for

- a device-code sign-in
- a service-principal route
- release notes per version
- a named support channel
- Publish the service tool schemas
- Add a changelog
- dedicated Entra app
- read-only mode

## When it's done

Send what changed and where it's published as a dispute (https://www.anchorterminal.com/builders/#disputes, or `POST https://www.anchorterminal.com/api/v1/contact` with `"kind": "dispute"`). Disputes are answered in public, and the listing is checked again by the same checklist. Paying for an audit or a listing claim changes nothing here.