mcp-discovery-surveyworkflow★0
caiolea0/hermes-cloud-studio ↗What it does
Discover and classify MCP ecosystem candidates for integration
Best for
Deterministic monthly/quarterly MCP ecosystem surveys that feed implementation planning without re-discovering
Inputs
- · optional: dimension cap, module cap overrides
Outputs
- · >=15 MCP candidates {name, repo_url, maintainer, category, fit_score}
- · JSON snapshot + MCP-DISCOVERY-RUN-{date}.md report
- · SQLite append-only mcp_discovery_runs DB row
Requires
- · Ollama (qwen2.5-coder:7b local)
- · git
- · SQLite
- · Bash/Glob for repo discovery
Preconditions
- · Ollama running on localhost
- · mcps/discovery/ directory writable
- · mcp_discovery_runs.db initialized
Failure modes
- · Sanity gate fails if <15 candidates collected (hard fail)
- · Ollama classify misses real fit (low precision)
- · Duplicates bypass dedup (false low count)
Trust signals
- · 8-phase deterministic pipeline (Bootstrap → Fan-out Search → Dedup → Classify → Rank → Critic → Persist)
- · Ollama classify with rigid schema (tools, fit_hermes, risk, effort)
- · Append-only DB preserves history
- · Hard sanity gate on minimum candidate count