100 code MCP servers. Those with live runtime data are ranked first by behavioral trust score (success rate, latency, uptime, per-reporter influence capping — not GitHub stars); servers not yet observed answering requests are shown UNRATED rather than given a fabricated score.
Category behavioral norms (across 16 measured servers): median latency 219ms · median success rate 0% · median trust score 37.2 — raw baselines.
Before your agent invokes any of these, run a free pre-flight check: GET /v1/preflight?target={server_url} — it returns a verdict and a signed diligence receipt. Or gate calls automatically with the trust-check SDK (PyPI: dominion-trust-check; hosted JavaScript module).
FAQ
What is the best code MCP server in 2026?
Based on independently measured runtime behavior, @contextgate/mcp currently ranks #1 in the code category with a trust score of 40.4/100 across 1 tracked interactions.
How are code MCP servers ranked?
Rankings use Dominion Observatory's behavioral trust score: real agent-reported success rates, latency distributions, and uptime — with per-reporter influence capping so no single source can dominate a score. Servers with no runtime data are UNRATED and excluded, never given a fabricated score.
How many code MCP servers are tracked?
Dominion Observatory tracks 18691 servers in the code category; the 100 with behavioral trust data are ranked here. Check any server before invoking it at /v1/preflight.