Agentic Resource Discovery Server

Agentic Resource Discovery Server

Turn your Obsidian vault into a local-first Agentic Resource Discovery (ARD) publisher and Agent Registry. The plugin scans your AI Skills, builds a rich ai-catalog.json, and runs a localhost server so AI agents on your machine can discover and fetch exactly the skill they need — by natural-language search — without those skills ever leaving your computer.

Built for people who have accumulated dozens or hundreds of AI Skills and MCP servers and don't want to load them all into every agent's context window. Instead of registering everything everywhere, publish them to a local catalog and let agents search it.

What it does

• Scans your skill folders (Anthropic Agent Skill format — SKILL.md + frontmatter) at startup, without blocking Obsidian, and turns each skill into a rich catalog entry: description, tags, capabilities, and synthesized example queries.
• Scans your subagent folders too (opt-in): each <name>.md definition with frontmatter (name, description, tools, model) and a system-prompt body — the Claude Code .claude/agents/ shape — becomes an entry of type application/ai-agent+md, so an agent can find "the editor" the same way it finds a skill.
• Serves an ARD registry over http://127.0.0.1 with a required bearer token:
  • GET /.well-known/ai-catalog.json — the public catalog
  • POST /search — natural-language search, results ranked 0–100 by relevance
  • POST /explore — facet counts (types, tags, capabilities) so an agent can see what's available before searching
  • GET /agents — deterministic, paginated listing, filterable by type, tags, and capabilities
  • GET /skills/<name>/SKILL.md (and bundled assets) — so an agent can fetch a skill's body and resources directly
  • GET /subagents/<name>.md — a subagent definition's body
  • GET /status — catalog counts per family, active search backend, and whether semantic embeddings are built yet
• Exposes an MCP endpoint (POST /mcp) using the Code Mode pattern: search, get_resource (get_skill remains as an alias), and execute tools, where execute runs sandboxed JavaScript against the catalog so an agent can filter and aggregate in a single call — with the same ranking as POST /search.
• Shows what's running. The settings Status panel reports the server URL, catalog size, last scan, embedding state, and the MCP endpoint — with one-click Copy MCP config and Copy curl example buttons.
• Stays fast on repeat use. Rescans only re-parse SKILL.md files that actually changed, and embedding vectors are cached across reloads, so a warm restart with a semantic backend is ready immediately.
• Adds zero mandatory downloads. The default search backend is an in-process BM25 index (MiniSearch) — no model, no network. Optionally, point it at a local embedding server you already run (Ollama, LM Studio, …) for hybrid semantic search; nothing is bundled or downloaded by the plugin, and it falls back to lexical if the server is down.

Status

Early but functional. The REST registry, skill scanning/enrichment, skill file serving, the MCP Code Mode endpoint, and optional semantic search (local or hosted embedding server) all work and are covered by 200+ tests. See the technical docs for the architecture and design.

isDesktopOnly — the plugin needs Node's HTTP server and filesystem access.

Install (manual / pre-release)

This plugin isn't in the community catalog yet. To try it:

1. Build it: bun install && bun run build (see DEVELOPMENT.md).
2. Copy manifest.json, main.js, and styles.css into <vault>/.obsidian/plugins/agentic-resource-discovery-server/.
3. Enable Agentic Resource Discovery Server in Obsidian → Settings → Community plugins.

Quick start

1. Open the plugin settings.
2. Under Skill folders, add one or more folders that contain skills (each skill is a subfolder with a SKILL.md). Folders may live outside the vault.
3. Optionally, under Subagent folders, add a folder of agent definitions (e.g. .claude/agents). Everything listed becomes part of the public catalog, so it is off until you add one.
4. Click Rescan now. The Status panel shows how many skills and subagents were indexed.
5. Copy the bearer token from the Server section (or use Copy MCP config in Status to get a ready-to-paste client config).
6. Point an agent at the registry:

# The public catalog needs no auth:
curl http://127.0.0.1:27182/.well-known/ai-catalog.json

# Search needs the bearer token:
curl -X POST http://127.0.0.1:27182/search \
  -H "Authorization: Bearer <your-token>" \
  -H "Content-Type: application/json" \
  -d '{"query":{"text":"summarize a long note"}}'

To use it as an MCP server, point your MCP client at http://127.0.0.1:27182/mcp with the bearer token. See the user guide for details.

Documentation

• User guide — usage, configuration, tips.
• Technical docs — architecture, domain model, and business rules.
• Contributing · Development
• What's new after updates. After a plugin update, a one-time dialog shows the release notes you just received (including skipped versions) with ways to support development. Never shown on fresh installs or regular restarts.

Privacy & security

Everything stays on your machine: the server binds to 127.0.0.1 only, every endpoint except the public catalog requires a bearer token (kept, like any API key, in Obsidian's secret storage on each device, never in the synced plugin data), skill file serving is confined to your configured folders (path-traversal-safe), and the execute sandbox has no network or filesystem access. No telemetry, no cloud.

License

MIT — by Sébastien Dubois. If this is useful, you can buy me a coffee ☕.

My other Obsidian plugins

Everything I build is documented in my newsletter and on my YouTube channel.

News & support

To stay up to date about this plugin, Obsidian in general, personal knowledge management and note-taking:

• Subscribe to my newsletter
• Subscribe to my YouTube channel
• Join the Knowii community and learn to organize your notes and put your knowledge to work, together with fellow knowledge workers

If this plugin is useful to you, here are the best ways to support my work ❤️:

• Join the Knowii community
• Become a GitHub Sponsor
• Buy me a coffee
• Subscribe to my YouTube channel
• Check out my products

Found a bug or have an idea? Open an issue.