The MCP Server Directory
Every Model Context Protocol server, scored on reliability, documentation, maintenance, and adoption. Independent, opinionated, refreshed daily from the official registry.
What this is
The MCP Server Directory is a live, independently maintained index of Model Context Protocol servers published to the official registry. Each server is scored 0–10 on four dimensions: reliability, documentation quality, maintenance velocity, and adoption. The directory is free, updated daily, and built by MegaOne AI.
Connect
3.9/10The MCP face of the hosted OAuth bridge — agents upload to YouTube through a human-granted
Connect Beyondcompare
3.9/10Mandate-bound compare/merge custody for agents: signed path grants, hash-chained compare packs.
Consent
3.9/10per-agent, per-purpose consent receipts for personal data — a grant an agent must present, and a
Countermark
3.9/10One verification call for third-party agent cards and payment mandates — a receipted verdict a
Covenant
3.9/10Consent-bound lead work for real-estate agents — provenance on every lead, a Gate on every send, a
Crosswalk
3.9/10Multi-regime AI-obligation mapping — which AI laws bind your systems, what evidence satisfies each.
Deepfield
3.9/10The governed database layer for agents — every store in one look, under policy: read-only by
Docket
3.9/10Agent-to-agent dispute process — structured claims, evidence packs, recorded determinations. Not a
Environment
3.9/10The smart-environment module — scopes an agent's authority to actuate a home, tier by tier, and
Envoy
3.9/10The signed front door for agent-to-agent business: vetted in, tier-told, drafted — never bound
Fathom
3.9/10DeFi, measured honestly: signed protocol registry, dual-source risk reads, and an allowance audit
Fedscope
3.9/10Federal + state opportunity search with fit and newcomer scoring and plain-language summaries — not
Florida Bid Intel
3.9/10Florida bid intelligence with honest deadlines — real competitions, real due dates, gaps stated not
Forecast
3.9/10The Polymarket adapter: the venue's prices, with their criteria and liquidity attached.
Haven
3.9/10The governed Zendesk server: OAuth-first, drafts until approved, and a receipt for every touch.
Health Check
3.9/10Scan an MCP server or agent configuration for security posture — evidence-bound findings
Hostbridge
3.9/10The sanctioned path for agents to deploy websites to shared hosting — scoped deploys, snapshots
Keyhold
3.9/10credential leases, never raw keys — scoped, expiring secret leases for agents; the agent gets a
Ledger
3.9/10Eclipse Ledger — betting records, price comparison, and responsible-gambling controls. Not a
Lineage
3.9/10spawn trees, model/config lineage, and version ancestry for agents — who spawned whom, running
Governed agent email identity — review-first sending, signed envelopes, and a fraud firewall that
Marquee
3.9/10The live-events agent: it finds the event, watches the price, buys only inside your mandate, and
Meridian
3.9/10Governed telehealth operations — jurisdiction-gated encounters, signed-note integrity, PHI access
Meter
3.9/10Cross-provider agent spend analytics — one ledger for tokens, tools, and payments.
Orbit
3.9/10The HubSpot wrap: the vendor's server, under policy — classified tools, approval gates, masking
Pantry
3.9/10The unattended-retail module — reads vending fleets' own telemetry and predicts stock-outs. It
Procure
3.9/10Intent to receipt in one fail-closed pipeline — policy before payment, abort on drift.
Proof
3.9/10Governed video production — Herald-signed approval pins over exact edit states, hash-pinned
Proving
3.9/10Governed bench access for agents — signed mandates, pinned artifacts, evidence for every session.
Revoke
3.9/10signed revocation lists for agent cards, keys, and mandates — one feed, every registry, revocation
Section508
3.9/10WCAG + Section 508 audits that produce the deliverable federal buyers actually require
Stage
3.9/10The console bridge — deploy and playtest AI-built games on your own Dev Mode console, with receipts.
Steward
3.9/10Team-chat agent governance — the signed mandate, verified disclosure, review-first posting and
Strike
3.9/10The Coinbase exchange adapter: spot products and books, normalized into the suite's one book.
Tether
3.9/10Governed robot task authority — signed mandates, bounded envelopes, evidence for every task.
Thread
3.9/10On-prem bridges for the legacy software-engineering toolchain, plus the Evidence Weaver that joins
Till
3.9/10Bazaar checkout for agents — quote, mandate, signed entitlement, delivery, receipt, refund.
Tripwire
3.9/10Runaway enforcement — spawn, loop, and spend-velocity detection with automatic mandate freeze.
Umbra
3.9/10The closed-vocabulary witness: cross-registry lookalike and endpoint-change watch, findings as
Wageward
3.9/10Payroll approval-and-evidence rail for agents — approve the exact run, or it refuses.
Zenith
3.9/10The fitness module of Eclipse Procure — reads the machine, turns the dial under a safety ladder
Zephyr
3.9/10The Bluesky adapter: real public reads over the open AT Protocol, with review-first and house caps
Clipai MCP
3.9/10Turn long videos into scored, captioned 9:16 clips and schedule them to social platforms.
Basiiik
3.9/10Design system for coding agents: component specs, rules, tokens, icons and checked React snippets.
MCP Server
3.9/10Check Excel workbooks locally; calculate them through the WorkbookAPI hosted API.
Crosslist
3.9/10Publish once, tracked everywhere: registry-native facets plus a mandate-bound listing ledger.
MCP Server
3.9/10Enterprise context layer for AI across 50+ work apps. Enterprise-only: book a demo for access.
Porticol MCP
3.9/10Coding agents call your APIs without the keys. A person approves risky calls; each call is logged.
Free · No spam
New servers, score changes, breakages
Every Monday: what shipped, what broke, what's worth trying. Five minutes. No filler.
Four dimensions, weighted
- Reliability (30%) — repository present, permissive license, not archived, official status.
- Documentation (20%) — description length, website presence, README quality.
- Maintenance (25%) — how recently the code was pushed; versioning discipline.
- Adoption (25%) — GitHub stars (log-scaled), remote availability, official endorsement.
What is MCP?
MCP stands for Model Context Protocol. It is an open standard introduced by Anthropic in late 2024 that lets AI assistants like Claude connect to external tools, data sources, and services through a common interface. An MCP server is a small program that exposes one capability (file access, a database, an API) to any MCP-compatible AI client.
Where does the data come from?
Every server in this directory is pulled from the official Model Context Protocol Registry at registry.modelcontextprotocol.io. We also enrich each entry with GitHub metadata (stars, last commit, license, primary language) where a repository is linked. The full re-scan runs daily.
How are servers scored?
Each server gets a 0–10 overall score, weighted from four sub-scores: reliability (30%), documentation (20%), maintenance (25%), and adoption (25%). The full methodology is visible on this page under "How scoring works".
Is this affiliated with Anthropic?
No. MegaOne AI is independent. Anthropic created the MCP standard and maintains the reference implementations, but this directory is built separately and rates all servers including Anthropic's own.
How often does the directory update?
The pipeline re-fetches the full registry and re-scores every active server once per day. Individual entries are refreshed whenever a new version is published to the official registry.
Is it free?
Yes. Browsing, filtering, and searching the directory is free. The weekly email briefing is free. There is no account required.