Repository navigation
📌 Update, Oct 2026: how onboarding works (community notes)This thread became the place to request listing on github.com/mcp. These notes collect what's been said here. The process (based on the accepted answer, @trent-j, May 2026):
Include all of these in your request:
Check before you post (read the values from the registry, not from memory): Check whether you're already listed on GitHub's side: Community notes from this thread. Corrections welcome. Original question (Oct 2025): Hi community! Looking for clarification on how the GitHub MCP Registry works. Server: claude-faf-mcp Questions Is github.com/mcp a curated list (vs automatic sync)? Why We Ask Previously posted: modelcontextprotocol/registry #690 Thanks for any insights! 🙏 |
Replies: 194 comments 10 replies
|
39 days, no response. claude-faf-mcp is approved by Anthropic (PR #2759 MERGED), has 10,067+ npm downloads, but isn't visible on github.com/mcp despite being listed in modelcontextprotocol/servers. What are the curation criteria for github.com/mcp? Is there a separate approval process beyond the Anthropic PR? Links: |
|
🎉 Update: Additional Milestone Since Posting
This makes claude-faf-mcp:
Full details: https://www.faf.one/blog/iana-registration Still curious about the github.com/mcp curation process and happy to |
|
Hey @Wolfe-Jam, |
|
Hey HolgerMany thanks for your reply, we will take of it and let you know. thanks again,James (Wolfe) HarrisonFounderWolfe James LLCfaf.one - Project DNA for AI contextIANA Registered format .FAF for AI context application/vnd.faf+yaml"README evolution for AI era"Google Gemini CLI(9.5/10 Rating)
On Nov 10 2025, at 8:52 am, Holger Staudacher ***@***.***> wrote:
Hey @Wolfe-Jam,
right now the GitHub MCP registry is a curated list. This will change in a couple days when we enable sync with the open source registry. In order to get your server added, you have to publish it to the open source registry first. Here is a guide how to publish your server if you haven't done it already.
—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you commented.Message ID: ***@***.***>
|
|
Hey @hstaudacher - following up from November. Just published both FAF packages to the open source registry today:
Ready for sync whenever that goes live. If there's anything blocking the registry sync rollout, happy to help test, validate, or contribute. We've been through the mcp-publisher flow and can provide feedback if useful. |
|
Hey @hstaudacher — friendly follow-up from the November conversation about registry sync. Quick update on where things stand: we now have 6 FAF MCP servers published to the open source registry, all passing
All published via You mentioned back in November that github.com/mcp would start syncing from the open source registry. Is that sync live now? If so, is there something on our end we need to do to appear? If it's still in progress, totally understand — just want to make sure we're not missing a step. Happy to help test or validate anything on the sync pipeline. |
|
Status check, 2026-05-18: github.com/mcp still shows ~97 servers, so the open-source-registry sync doesn't look live yet. The FAF servers remain published & schema-clean on the OSS registry. If the list stays curated long-term: Is there an ETA for the sync, or is github.com/mcp staying curated? Either answer's fine — closing the loop on this thread. |
|
Hi @Wolfe-Jam, thanks for the thoughtful follow-up and for continuing to publish and validate your servers. A few clarifications from our side: the GitHub MCP Registry now has the ability to sync versions from the open source registry, but onboarding a new server is still a manual curation process today. Once a server has been onboarded, newly published versions from the OSS registry should sync from there. I’m closer to the technical side than the product/curation criteria, so I don’t want to overstate anything here. I’ll share your request with our product team and ask whether we’re able to move forward with onboarding them. Thanks again for your patience, and for offering to help validate the sync path. |
|
Hey @trent-j — thanks again for looking into this and for offering to take it to the product team, genuinely appreciate it. Quick update from our side: since May we've moved our two flagship servers to the FAF Foundation's DNS-verified No rush at all on the github.com/mcp side — mostly I'd love to understand the lay of the land so we can be ready: is onboarding/sync something that's moving, or staying curated for now? And if there's a process, criteria, or anything we should prep on our end to make it easy when the time comes, happy to do the legwork. Either way — thanks for keeping the channel open. 🙏 |
Onboarding request:
|
| Field | Value |
|---|---|
| Official registry name | io.github.edithatogo/fyi-mcp |
| Latest version | 0.1.2 (active / isLatest) |
| Repository | https://github.com/edithatogo/fyi-cli |
| server.json | https://github.com/edithatogo/fyi-cli/blob/master/server.json |
| Description | Local-first FOI/OIA request tracker MCP for Alaveteli platforms (FYI.org.nz and multi-jurisdiction catalog) |
OSS registry verification
curl -sS "https://registry.modelcontextprotocol.io/v0/servers?search=fyi-mcp"
returns io.github.edithatogo/fyi-mcp v0.1.2 active/latest.
github.com/mcp status (2026-07-09)
- Search
fyi-mcp/edithatogodoes not surface a curated card for this server https://github.com/mcp/io.github.edithatogo/fyi-mcp→ 404
Could the product/curation team please consider onboarding this server so OSS version sync can take effect? Happy to provide additional metadata or validation help.
Tracked in: edithatogo/fyi-cli#32
Onboarding request:
|
Onboarding request:
|
| Field | Value |
|---|---|
| Official Registry name | io.github.happy520ai/unified-ai-system |
| Latest version | 0.4.8 (active) |
| Package / transport | ghcr.io/happy520ai/unified-ai-system/mcp-server:0.4.8 / stdio |
| Repository | https://github.com/happy520ai/unified-ai-system |
| Manifest | https://github.com/happy520ai/unified-ai-system/blob/master/server.json |
| Documentation | https://happy520ai.github.io/unified-ai-system/codex-mcp-docker-quickstart.html |
Unified AI System is an Apache-2.0 self-hosted gateway that gives Codex, Cursor, Cline, and generic MCP clients nine bounded tools. Its differentiating first-run path turns rough natural-language requests into structured, reviewable prompts locally, with a deterministic fake provider and no account or API key required. Real provider calls remain explicit and disabled by default.
The repository publishes pinned OCI images and runs public CI, container, MCP discovery, security-scanner, clean-clone, and process-cleanup checks. The Official Registry currently contains versions 0.3.1 through 0.4.8; GitHub MCP search still returns no result.
Please consider onboarding this server so future Official Registry versions can sync. I am happy to provide any additional metadata or validation evidence the curation team needs.
|
Onboarding request: Following the OSS-registry-first process confirmed in this thread (publish to the open source registry, then request manual onboarding; versions sync after):
Derive gives agent output a permanent home. One call publishes a document, styled page, deck, or report to a versioned URL a teammate can open and comment on inline; the agent reads those comments back and revises. Every tool carries a title and honest Nothing to install and no key to paste — it is a remote server behind OAuth with dynamic client registration. Self-hosting is a single container, for teams that cannot send work to a hosted service. OSS registry verification: returns github.com/mcp status (2026-08-12): a search for Could the curation team consider onboarding this so future OSS registry versions sync automatically? Happy to provide any further metadata or validation. |
|
Onboarding request for a server already published to the OSS registry, following the format used earlier in this thread.
Happy to provide a populated test account for review. Thanks for considering. |
|
Requesting onboarding for a new server:
OSS registry verification: curl -s "https://registry.modelcontextprotocol.io/v0/servers?search=io.github.rafaelgaspar/longhorn-mcp"returns the server with Publishing is automated via GitHub Actions OIDC on every tagged release ( |
|
Please consider AIHOT for initial onboarding to github.com/mcp. AIHOT helps users read and search curated AI news in Chinese, follow multi-source hot topics, and get daily, weekly, and monthly briefings with source links.
Verified on 14 October 2026: the server exposes eight read-only tools. Latest-news, keyword-search, and daily-briefing calls succeeded both directly and through Grok's custom MCP connector. GitHub's registry search currently returns no AIHOT entry. Verification: https://api.mcp.github.com/v0.1/servers?search=aihot Please onboard the existing registry identity so future versions can sync. Thank you! |
|
Onboarding request: Senzing ( Published to the official MCP Registry (DNS-verified
What it does: entity resolution for Identity Intelligence — dedupe, match and link people and company records across systems. The server grounds an assistant in Senzing's own documentation and indexed code (data mapping to the Senzing Entity Specification, SDK code for Python/Java/C# plus community Rust/TypeScript, error troubleshooting, sample data). It needs no live Senzing instance and sends no user data to Senzing; mapping and resolution run locally. Also listed in the Claude Connectors Directory. |
Onboarding request:
|
| Field | Value |
|---|---|
| Registry name | ai.khive/khive |
| Version | 1.0.0, active, isLatest: true |
| Transport | Remote, Streamable HTTP at https://khive.ai/v1/mcp |
| Auth | OAuth. An unauthenticated request gets 401 with WWW-Authenticate pointing at the protected-resource metadata (authorization server https://auth.khive.ai, scope khive:mcp). Clients without OAuth can send an API key as Authorization: ApiKey <key>. |
| Repository | Hosted service. The open-source engine is ohdearquant/khive. |
| License | Engine: Apache-2.0. The hosted service is a paid plan. |
| What it does | Persistent memory, a knowledge graph and a task list for agents, kept in the user's own workspace across sessions and clients. 22 tools, each with a title and a read-only or destructive hint. |
| Docs | https://www.khive.ai/docs |
Check:
curl "https://registry.modelcontextprotocol.io/v0.1/servers?search=ai.khive/khive"
Leo, khive AI
|
Please consider Hoot Lens for initial onboarding to github.com/mcp. We publish the existing Hoot Lens gives coding agents recorded website evidence: timed visit narratives, click maps, recurring snags and linked JavaScript errors. Agents can trace a finding to a page, release, device and recorded moment before proposing a fix.
The release record links verified npm signature/provenance evidence to public source Verification at 4:28 PM Pacific on 14 October 2026: curl 'https://registry.modelcontextprotocol.io/v0.1/servers/com.hootlens%2Fmcp/versions/latest'
curl 'https://api.mcp.github.com/v0.1/servers?search=hootlens'The upstream entry is active/latest 1.1.0. GitHub's API returned zero entries for Please consider onboarding this identity so future versions can sync. Happy to provide additional metadata or reviewer instructions. Thank you. |
|
Please onboard RealtyCrawl to the curated GitHub MCP Registry.
Five tools discover platforms and current capabilities, inspect endpoint schemas, and execute explicitly budgeted operations for source-linked real estate observations. Discovery includes research/unavailable operations: check capabilities before execution; it is not a guarantee of live source coverage. Metadata is free. Verification: curl 'https://registry.modelcontextprotocol.io/v0.1/servers?search=io.github.realtycrawl%2Frealtycrawl'Registry readback returned version 1.0.0 with status |
Onboarding request: ai.glasser/glasserPlease consider Glasser for the curated GitHub MCP Registry. The server is already published in the official MCP Registry.
Glasser lets AI agents discover specialist data APIs, inspect endpoint contracts and prices, and retrieve data through one account and balance. Data retrieval is paid per call. Registry verification: https://registry.modelcontextprotocol.io/v0.1/servers?search=glasser The GitHub MCP Registry search returned no matching Glasser server today. Thank you for considering this onboarding request. |
|
Onboarding request: Watchgoose
Verification Thanks! |
|
Onboarding request: 21RISK ( Published to the official MCP Registry first, following the process in this thread. Values read from the registry on 2026-10-15.
A 21RISK account is required; a free account is enough to try it. Verification: |
Onboarding request:
|
| Field | Value |
|---|---|
| Registry name | com.artifactbridge/artifactbridge (domain-verified namespace) |
| Version | 0.1.0, active, isLatest: true |
| Transport | Remote Streamable HTTP: https://app.artifactbridge.com/mcp |
| Auth | OAuth 2.1 with PKCE (S256) and dynamic client registration (RFC 7591). No API key. |
| Repository | https://github.com/omnim-ai/artifactbridge-plugin (public plugin and skills for the server) |
| License | MIT (plugin repository) |
| Website | https://www.artifactbridge.com |
What it does: ArtifactBridge gives coding agents governed access to a team's documents. Agents search, read, comment, and propose changes. A human approves every change before it is published.
Verification:
curl -s "https://registry.modelcontextprotocol.io/v0.1/servers/com.artifactbridge%2Fartifactbridge/versions/latest"Thank you.
Onboarding request:
|
| Field | Value |
|---|---|
| Registry name | com.opentext/sdp-remote-mcp (domain-verified namespace) |
| Version | 1.0.0, active, isLatest: true |
| Transport | Remote Streamable HTTP: https://almoctane-{region}.saas.microfocus.com/mcp |
| Auth | OAuth 2.1 (interactive browser flow) · API access key (headless or service use) |
| Repository | https://github.com/opentext/OpenText-Model-Context-Protocol-Server |
| What it does | Connects the AI client to your OpenText Core Software Delivery Platform workspace and interact with delivery data without leaving the IDE. |
| License | OpenText End User License Agreement |
| Website | OpenText-Model-Context-Protocol-Server |
Verification:
curl https://registry.modelcontextprotocol.io/v0.1/servers?search=com.opentext/sdp-remote-mcpThank you.
|
Onboarding request: ai.memestack/mcp (MemeStack)
|
Onboarding request:
|
| Field | Value |
|---|---|
| Registry name | sh.openpath/openpath |
| Version | 3.6.0, active, isLatest: true |
| Remotes | https://openpath.sh/mcp/search (Streamable HTTP, no authentication) and https://openpath.sh/mcp (OAuth, for messages and the user's own listings) |
| Tools (no-auth endpoint) | search, get_listing, both read-only |
| Homepage / docs | https://openpath.sh, https://openpath.sh/docs/connect |
Openpath is an open directory of real-world listings for AI agents: live jobs from employers' own job boards (Austin, TX and remote), each with the employer's Apply link, and local businesses around Austin, each with an Openpath inbox. Results are plain Markdown, one listing per result. Also listed in the Claude connectors directory, Smithery and Glama.
Onboarding request:
|
| Field | Value |
|---|---|
| Registry name | com.liquidcompute/gcat (domain-verified namespace) |
| Version | 1.0.0, active, isLatest: true |
| Transport | Remote Streamable HTTP: https://api.liquidcompute.co/mcp |
| Auth | OAuth 2.1 with dynamic client registration and client ID metadata documents, scope gpu_tests. An unauthenticated request gets 401 with WWW-Authenticate pointing at the RFC 9728 protected-resource metadata. No API key. |
| Repository | https://github.com/Liquid-Compute/gcat (listing files and agent skill; the server is hosted) |
| License | MIT for the repository; the hosted service is covered by https://liquidcompute.com/terms |
| Website | https://liquidcompute.com/gpu-test |
| Pricing | Free; needs a Liquid Compute account (public sign-up) |
What it does: runs GPU acceptance tests on NVIDIA GPU nodes the user administers. Liquid Compute connects over SSH from its own servers, runs NVIDIA DCGM diagnostics, a CUTLASS GEMM benchmark, HBM and PCIe bandwidth tests and NCCL all-reduce, and reports each measurement against the GPU profile's thresholds (A10, H100, H200, B200, B300; single nodes and 2- or 4-node RDMA clusters). Nothing is installed on the node. Adding nodes and starting a test require the user's explicit confirmation, enforced server-side. 10 tools, all annotated (read-only, or destructive for start and stop).
Verification:
curl "https://registry.modelcontextprotocol.io/v0.1/servers?search=com.liquidcompute/gcat"
curl "https://api.mcp.github.com/v0.1/servers?search=liquidcompute" # not listed yet
Onboarding request:
|
| Field | Value |
|---|---|
| Registry name | com.makepartsnow/manufacturing (HTTP-verified namespace on makepartsnow.com) |
| Version | 1.0.0, active, isLatest: true |
| Transport | Remote Streamable HTTP: https://makepartsnow.com/mcp |
| Auth | Optional API key, Authorization: Bearer mpn_agent_.... initialize, tools/list, the catalog and sign-up work without it; the create_agent_account tool returns a free key for quoting and ordering. |
| Repository | None: hosted, closed-source service. Public listing packages: https://github.com/makepartsnow/makepartsnow-claude |
| License | Proprietary hosted service; Terms, Privacy |
| Website | https://makepartsnow.com/developers |
What it does: instant quotes and ordering for custom parts from CAD files: SLS nylon and FDM 3D printing, and CNC-machined 6061 aluminum or C360 brass, made in-house and shipped within the US. 18 tools; a person pays through a Stripe Checkout link, or an agent pays in USDC via x402.
Verification:
curl "https://registry.modelcontextprotocol.io/v0.1/servers?search=com.makepartsnow"
curl "https://api.mcp.github.com/v0.1/servers?search=makepartsnow" # not listed yet
Onboarding request: io.github.aibakery/houstonPlease consider Houston for the curated GitHub MCP Registry. Houston lets agents read, write, and combine connected apps and databases through a shared runtime, with reusable workspace code and per-person permissions. Connector credentials stay on Houston.
Verification: curl 'https://registry.modelcontextprotocol.io/v0.1/servers/io.github.aibakery%2Fhouston/versions/latest'
curl 'https://api.mcp.github.com/v0.1/servers?search=io.github.aibakery%2Fhouston'The upstream entry is active/latest. An unauthenticated initialize request returns HTTP 401 with a WWW-Authenticate pointer to https://ok-houston.com/.well-known/oauth-protected-resource/mcp. Authorization metadata is available at https://ok-houston.com/.well-known/oauth-authorization-server. Houston is an early-stage product; no uptime SLA or broad production adoption is claimed. |
Hi @Wolfe-Jam, thanks for the thoughtful follow-up and for continuing to publish and validate your servers.
A few clarifications from our side: the GitHub MCP Registry now has the ability to sync versions from the open source registry, but onboarding a new server is still a manual curation process today. Once a server has been onboarded, newly published versions from the OSS registry should sync from there.
I’m closer to the technical side than the product/curation criteria, so I don’t want to overstate anything here. I’ll share your request with our product team and ask whether we’re able to move forward with onboarding them.
Thanks again for your patience, and for offering to help valid…