M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
D

Delimit MCP Server

Community
The merge gate for AI-written code, with signed, replayable attestation.
How to install ↓ Compare GitHub source repository ↗
Category
Dev Tools #280 of 438
Stars
★ 24 Popular
Transport
stdio (local process)
Runtime
Node.js
Credentials
No credential needed
License
MIT
Last commit
Tools
13
48FMRS · D

Delimit MCP server provides a comprehensive AI governance toolkit, suitable for teams wanting to enforce policies and maintain audit trails across multiple models. It has a generous free tier and a rich feature set, but requires careful permission and risk management.

Strongest · Documentation 15/20 Weakest · Reliability 6/20

Reliability
6/20
Security and permissions
8/20
Maintenance
9/20
Documentation
15/20
Setup experience
10/20
Why each score
Reliability 6/20
Based on the repository metadata and README, the server is delivered via the npm package delimit-cli over stdio. Actual tool listing and implementation are not visible in static files; the README claims 4800+ passing tests, but no test files or CI config are attached for verification. The repo is un-archived with zero open issues and version 4.18.8, suggesting activity; however, there is no executable evidence that the server starts cleanly and completes the MCP handshake. Static calibration caps reliability at 12, but a lack of direct execution evidence reduces the score to 6.
Security and permissions 8/20
No immediate red flags are present. The README offers extensive transparency about telemetry and cloud sync: local-by-default storage, optional Supabase sync requiring explicit credentials, a kill-switch environment variable, and no hardcoded secrets. Confirmation mechanisms for evidence collection and seal verification are described, with a least-privilege adoption path recommended. No malicious behavior or dangerous defaults surface. Although network behavior is well-documented, it cannot be verified from static files alone; therefore, deductions are applied, yielding a score of 8.
Maintenance 9/20
The repository is actively maintained (un-archived, version 4.18.8, 21 stars), carries an MIT license, and includes a clear roadmap with entitlements such as secret management and SLSA provenance on npm. No open issues are visible. Sustained maintenance evidence is limited—commit history, contributor details, and dependency update records are not part of this static review; however, the release history and versioning details in the documentation indicate ongoing development. Ownership, license, and update paths are transparent, and activity is moderate, supporting a score of 9.
Documentation 15/20
The README is extensive, covering installation, CLI commands, security disclosures, pricing tiers, telemetry and data privacy policies, upgrade notes, and an FAQ. The tool surface is enumerated with two tiers. Limitations include unclear gating for Pro-only features and the fact that some facts (like pricing) may change. Lacking full API reference or code samples, but overall documentation quality is high; hence, a score of 15.
Setup experience 10/20
Install instructions are succinct: run the CLI via npx, server is an npm package with a single 'delimit-mcp-server' binary. However, configuration examples for mainstream clients (Claude Code, Cursor, Gemini CLI) are absent. Additional steps may be required during runtime (e.g., npm install, environment configuration), and the MCP server rides on the entire CLI package, which could include unnecessary dependencies. Core steps are clear but lack specific cross-client configuration and setup automation. Given the static limitation, setup is scored 10.

Static review · not runListed 2026-09-08

Read the FMRS scoring method →

Fit and risk

What it can accessReads local filesWrites / deletes local filesRuns commands or codeUses the network

Best for

  • Teams using multiple AI coding assistants
  • Organizations wanting to enforce API compatibility
  • Compliance workflows needing verifiable audit trails of AI changes

Not for

  • Users not using AI coding tools
  • Teams seeking a full CI/CD replacement rather than a governance layer

Required permissions

  • Read repository files (for scanning, linting, diffing)
  • Write to local .git directory and .delimit directory (for hooks, configs, memory)
  • Execute commands (e.g., tests, security audit)
  • Network access (for model deliberation, cloud sync if enabled)

Risks and side effects

  • The server can execute code (e.g., tests, security audit), which could pose security risks.
  • Credentials or sensitive data could be stored in memory or ledger; ensure proper access control.
  • If cloud sync is enabled, event data may be sent to external services (user-provided Supabase).

Setup

Before you start

Runtime：Node.js

SUPABASE_SERVICE_ROLE_KEY optionalsecret Optional: Supabase service role key (a secret) needed only if you enable self-owned Supabase sync
Other optional settings (3)
DELIMIT_DISABLE_CLOUD_SYNC optional Optional kill switch; setting to 1 disables cloud sync for a shell session
SUPABASE_URL optional Optional: URL of your own Supabase project for opt-in cloud mirroring; unset means fully local
DELIMIT_WEBHOOK_URL optional Optional webhook endpoint for governance events; unset by default, so nothing is sent

Install the CLI globally via npm:

npx delimit-cli init

This sets up the merge gate configuration. No account or API keys required. For a quick zero-config check:

npx delimit-cli check

Check that it works

Run npx delimit-cli doctor, or check the client's tool list for delimit_scan, delimit_lint and other delimit_* tools; npx delimit-cli check also proves the CLI works with zero config.

Troubleshooting

  1. If the server does not start, check that Node.js is installed and version is sufficient.
  2. Run `npx delimit-cli doctor` for diagnostics.
  3. For tool call failures, ensure you are in the correct repository directory.

Things to try

Once connected, you can ask your AI assistant things like:

  • Diff old.yaml and new.yaml OpenAPI specs and list every breaking change
  • Run a lint check on the api/openapi.yaml change and give a pass/fail verdict
  • Classify this change as MAJOR/MINOR/PATCH and assess the blast radius for downstream callers
  • Remember that the v1 /users field is frozen until Q3 because billing depends on it, then recall the billing memories

Tools 13

delimit_scan read-only
Scans project to discover OpenAPI specs, frameworks, security issues, and tests.
delimit_init writes
Initializes merge gate configuration and policy presets.
delimit_lint read-only
Checks spec changes for breaking changes.
delimit_diff read-only
Compares two OpenAPI specs showing structural differences.
delimit_semver read-only
Determines semantic version change type (MAJOR/MINOR/PATCH/NONE).
delimit_impact read-only
Scans dependency manifests to assess impact scope of changes.
delimit_deliberate read-only
Runs multi-model deliberation between AI models to reach consensus.
delimit_memory_store writes
Stores persistent memories (context and decisions).
Show 5 more tools
delimit_memory_search read-only
Semantic memory search across sessions.
delimit_recall read-only
Recalls saved memories locally or with filters.
delimit_ledger_add writes
Adds task or item to shared ledger.
delimit_ledger_context read-only
Gets top open items by priority at session start.
delimit_ledger_done writes
Closes a task item with a note, optionally capturing PR URL as proof.

Use cases

Protect APIs from breaking changes before merge
Maintain context and decisions across models and sessions
Resolve difficult governance judgment calls via multi-model deliberation
Generate signed, replayable audit trails for compliance

Supported clients

Claude Desktop
Codex
Cursor
Gemini CLI

Listed from the project's documentation, not tested by this site.

Overview

Delimit is an MCP server that provides a governance chain for AI-assisted coding. It wraps various AI coding assistants (like Claude Code, Codex, Cursor, and Gemini CLI), runs gate checks, records changes, and signs replayable receipts for every merge. It offers API governance tools such as lint and diff detection of 28 change types (including 17 breaking changes), semantic version classification, multi-model deliberation, persistent memory and task ledger, and governance features like security audit and evidence collection.

Similar servers

Context7 80 · B

Upstash's official server providing up-to-date third-party library docs for AI coding assistants

★ 63.3k · Tools 2 Compare with this →
Compare with other MCP servers

Source revision d6f14b58f4f3 Data synced 2026-10-19 Read the FMRS scoring method