AI Workbench MCP
Wire Goose (and other MCP clients) to Workbench acceptance evidence, validation gates, and analytics so agents can prove features pass before you ship.
Overview
AI Workbench MCP is an MCP server for the Ship phase that exposes Workbench acceptance evidence, validation gates, and analytics to Goose and other MCP clients.
What is this MCP server?
- Goose-first MCP server for AI Workbench workflows
- Surfaces Workbench-owned acceptance evidence for agent-driven QA
- Exposes validation gates agents can check before merge or release
- Workbench analytics callable from the agent via MCP stdio
- PyPI package ai-workbench-mcp at version 0.6.0a0 with stdio transport
- Server schema version field: 0.6.0a0 (alpha)
- 1 PyPI package identifier: ai-workbench-mcp
- Transport: stdio
What problem does it solve?
Agents say work is complete without reading your real acceptance gates or evidence stored in Workbench.
Who is it for?
Indie builders using Goose or MCP-enabled agents alongside Workbench-style acceptance and gate workflows.
Skip if: Teams with no Workbench or acceptance model who only need a generic test runner MCP.
What do I get? / Deliverables
Your agent can query validation gates and acceptance evidence over MCP so release decisions track Workbench truth instead of chat optimism.
- MCP tools backed by Workbench acceptance evidence
- Queryable validation gate status for agents
- Analytics surfaced to the agent session
Recommended MCP Servers
Journey fit
Acceptance evidence and validation gates are the canonical home in Ship because they gate release readiness, even when you start capturing evidence during Build. Testing is where structured pass/fail gates and acceptance artifacts belong—not ad-hoc chat claims about done-ness.
How it compares
Workbench validation bridge via MCP, not a replacement skill or standalone CI product.
Common Questions / FAQ
Who is AI Workbench MCP for?
Builders who run Goose or other MCP clients and want agents to read Workbench acceptance evidence and validation gates before shipping.
When should I use AI Workbench MCP?
Use it when you capture acceptance in Workbench and need agents to verify gate status during testing and pre-release checks—not for one-off scraping or hosting.
How do I add AI Workbench MCP to my agent?
Install the PyPI package ai-workbench-mcp, add a stdio MCP server entry in Goose or your client config pointing at that package, then restart the agent so tools for evidence, gates, and analytics appear.