AgentRunDB

Know what actually works.

AgentRunDB tests AI agents, models, skills and MCP servers — and publishes reproducible compatibility data, real runs and working recipes.

Recently Verified

3 runs · Success rate 100% · Last verified 2026-08-19

5 runs · Success rate 80% · Last verified 2026-08-19

5 runs · Success rate 100% · Last verified 2026-08-19

1 runs · Success rate 100% · Last verified 2026-08-19

2 runs · Success rate 50% · Last verified 2026-08-18

Structured compatibility data, backed by real runs.

Does agent A actually work with model or tool B? Tested combinations with exact versions and evidence.

The underlying facts: every test run with environment, duration, cost and known issues.

Reproducible configurations backed by AgentRunDB runs — not AI-generated tutorials.

Reproducible configurations backed by real runs.

Which agents honor Agent Skills when running on DeepSeek — the current verified map.

● Backed by 0 AgentRunDB runs

Attach MCP servers (e.g. GitHub) to an agent that runs on DeepSeek. Configuration pattern for Claude Code — verification runs queued.

● Backed by 0 AgentRunDB runs

Point Claude Code at the DeepSeek API with three environment variables — the config our own CI runs on. AgentRunDB-verified.

● Backed by 2 AgentRunDB runs

Verified Runs

Compatibility Records

Agents

Skills & MCP Servers