Add AgentIndex MCP server — AI agent registry with 25K+ agents#1026
Add AgentIndex MCP server — AI agent registry with 25K+ agents#1026agentindexworld wants to merge 1 commit intostacklok:mainfrom
Conversation
JAORMX
left a comment
There was a problem hiding this comment.
Hey @agentindexworld, thanks for submitting AgentIndex! I went through the PR and the upstream repo. There are a good number of spec issues to work through, plus some things the maintainer team will need to discuss.
server.json spec issues
The server.json is pretty far from the expected schema. I'd suggest looking at an existing remote server entry like registries/toolhive/servers/stripe-remote/server.json as a reference. Here's what needs fixing:
$schemais missing — add"$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json"nameformat is wrong — should be"io.github.stacklok/agentindex", not just"agentindex"titleis missing — a human-readable display name like"AgentIndex"versionfield is wrong — use top-level"version": "1.0.0", not"version_detail": { "version": "..." }iconsarray is missing from top-level fieldsicon.svgfile is missing — needs to be included in the PR_metablock is entirely absent — this is wheretier,status,tools,overview, andtagsgo. The extension key inside_metamust exactly matchremotes[0].url(so"https://agentindex.world/mcp")remotes[0].transport_typeshould beremotes[0].type— the field name is justtypetoolsat top level — tools belong inside the_metaextension block, not at the JSON root. And they should be a flat string array (["check_agent", "register_agent", ...]), not objects with name/descriptiondescriptionexceeds 100 chars — needs to be trimmed down- Tags should include
"remote"
Repository assessment
The upstream repo (https://github.com/agentindexworld/agentindex-mcp-server) exists and has an MIT license, which is good. But it has only 1 commit and a single src/index.js file. No CI, no tests, no SECURITY.md, no dependency automation (Dependabot/Renovate), no releases. That's quite thin for our inclusion criteria.
Also note: the commit message lists get_autonomy_level as a tool, but the PR's server.json lists get_reputation instead. Worth double-checking which tools the server actually exposes.
Things the team needs to discuss
- Remote server to third-party infrastructure: All tool calls would go to
https://agentindex.world/mcp, which you operate. For remote servers we can't audit what's running on the other end, so we rely more heavily on the source code and project maturity... which is pretty early here. - Proprietary agent economy: The
$SHELLcurrency, Bitcoin-anchored identity proofs, and the registration model are a pretty unusual dependency to introduce into the catalog. We'd need to think about whether this fits the scope. - No auth configured: The server involves agent registration and identity management but there's no
oauth_configor authentication mechanism in the entry. That seems like a gap for a service handling identity.
What to do
- Rework
server.jsonfrom scratch using the stripe-remote entry as a template - Add the
icon.svgfile - Beef up the upstream repo (CI, tests, releases, SECURITY.md)
- Clarify the auth story for the remote endpoint
Happy to answer questions about the schema!
|
Hey @agentindexworld, thanks for taking the time to submit AgentIndex! We're going to close this one for now. The upstream project doesn't quite meet our registry inclusion criteria yet... the repo has a single commit and is missing CI, tests, releases, a security policy, and dependency automation. Those are things we look for to ensure the servers in the catalog meet a baseline of maturity and security. The If the project matures and you'd like to resubmit, we'd be happy to take another look! Check out the Thanks again for your interest! |
AgentIndex MCP Server
Global AI Agent Registry with RSA-2048 cryptographic passports.
Tools provided:
check_agent— Verify if an AI agent is registered and trustedregister_agent— Register a new agent and get a cryptographic passportfind_agent— Find agents to collaborate with by skillverify_passport— Verify a cryptographic passportget_reputation— Get full reputation and security reportStats: 25,000+ agents, 14 nations, autonomy levels 0-5
Links: