AI Workbench MCP
Goose-first MCP server for Workbench-owned acceptance evidence, validation gates, and analytics.
Open source Open in the app JSON README (API)
About
Goose-first MCP server for Workbench-owned acceptance evidence, validation gates, and analytics.
Details
- Kind
- MCP servers
- Topic
- Marketing & analytics
- Publisher
- hrishikesh-thakre
- Origin
- official
- Category
- ferramentas
- Transport
- local
- Version
- 0.6.0a0
- Last push
- 2026-06-27T17:03:45Z
- Repository state
- ativo
- Language
- Python
- License
- Apache-2.0
- Added
- 2026-08-29 04:00:08
- Updated
- 2026-08-29 04:00:08
- Origin id
io.github.hrishikesh-thakre/ai-workbench-mcp
README
# AI Workbench <!-- mcp-name: io.github.hrishikesh-thakre/ai-workbench-mcp --> [](LICENSE) AI Workbench supervises AI coding agents, captures evidence, validates work, applies acceptance policy, and produces auditable PR-ready reports. The PyPI package remains `ai-workbench-mcp` for this public alpha because the `ai-workbench` package name is already occupied. The product and CLI are **AI Workbench**: ```bash pip install ai-workbench-mcp ai-workbench --help ``` Current source metadata targets unpublished `ai-workbench-mcp==0.8.0a0`. This public alpha consolidates local supervision, evidence capture, validation, acceptance policy, and PR reporting into one product surface. ## Public Alpha Warning The supervisor is the preferred automated evidence path, but daemon, Codex hook, and OpenCode adapter coverage are alpha mechanisms. AI Workbench checks evidence quality and acceptance readiness; it does not prove the work is absolutely correct. High-risk work still requires human review. ## Architecture - AI Workbench supervisor captures local evidence. - AI Workbench validation writes `validation_report.json`. - AI Workbench quality gate writes `revision_decision.json`. - AI Workbench PR/report surfaces render `accept`, `needs_review`, or `block`. Agent output is a proposal. Workbench accepts evidence. MCP is the connection protocol. AI Workbench MCP is the tool server. Acceptance is decided by the selected validation profile and quality gate. The agent performs. Workbench accepts. MCP connects them. ## Quick Start Register a project once and start the local supervisor: ```bash pip install ai-workbench-mcp ai-workbench supervisor setup --project-dir . --task-type code_change ai-workbench supervisor start ``` Run Codex, OpenCode, Goose, or another supported local workflow in the project. Then inspect the latest report: ```bash ai-workbench supervisor status ai-workbench reports show latest --project-dir . ``` Render PR-ready artifacts from a finalized run: ```bash ai-workbench pr-gate --run-dir runs/<run_id> ``` The canonical local run ledger is: ```text runs/<run_id>/ task_metadata.json final_prompt.md model_selection.json model_output.md validation_report.json revision_decision.json run_log.jsonl metadata.json transcript.jsonl commands.jsonl workspace/ validation/ artifacts/ ``` `validation_report.json` and `revision_decision.json` are the final acceptance authority. Supporting supervisor reports are local evidence, not a substitute for those Workbench artifacts. ## Codex Hooks Install project-local Codex hooks: ```bash ai-workbench setup codex --project-dir . --task-type code_change ``` Restart Codex or start a new session, open `/hooks`, review the project hook, and trust it once. Until a hook event is observed, supervisor status reports Codex coverage as configured but unverified. ## Goose MCP AI Workbench still exposes the same MCP tool lifecycle. Register the server with Goose or another MCP host using: ```bash ai-workbench mcp serve ``` The seven MCP tools remain: ```text workbench_open_run workbench_select_policy_pack workbench_select_model workbench_record_execution workbench_validate_run workbench_quality_gate workbench_analyze_runs ``` ## PR Gate Workbench PR acceptance consumes real Workbench run evidence: ```bash ai-workbench pr-gate \ --run-dir runs/<run_id> \ --out runs/pr_gate/pr_comment.md \ --json-out runs/pr_gate/pr_decision.json ``` Outcomes are exactly: - `accept` - `needs_review` - `block` Missing, unreadable, or scaffold-only evidence blocks. A green CI run, uploaded artifact, sticky PR comment, or model self-claim is not acceptance evidence. ## Bootstrap Assets To add starter configs, prompts, recipes, docs, and the GitHub PR-gate workflow to a repository: ```bash ai-workbench bootstrap --target . ``` The bootstrap keeps `runs/` ignored. ## Package Demo For a package-only synthetic demo: ```bash ai-workbench demo --target ./workbench-first-run ``` This shows `accept`, `needs_review`, and `block` PR-gate outcomes with fixture evidence. It is not a real target-repository acceptance run. ## Development ```bash python -m pip install -e ".[dev,publish]" python -m pytest -q -p no:cacheprovider python -m ruff check . --no-cache python -m mypy --no-sqlite-cache --no-incremental ai-workbench demo --target runs/package_demo_smoke ai-workbench validate --project ai_workbench_mcp --profile scaffold --run-dir runs/scaffold-smoke ``` Do not commit `runs/`. Committed sample evidence must be sanitized and live under `examples/`. ## Docs - [Supervisor docs](docs/supervisor/automated-evidence-supervisor.md) - [Evidence folder contract](docs/supervisor/evidence-folder-contract.md) - [Transcript schema](docs/supervisor/transcript-schema.md) - [Workspace hygiene](docs/supervisor/workspace-hygiene.md) - [Confidence rules](docs/supervisor/confidence-rules.md) - [How acceptance works](docs/concepts/how-acceptance-works.md) - [Contract baseline](docs/contracts/v0.2-contract-baseline.md) - [Package demo walkthrough](docs/walkthroughs/package-demo.md) - [Codex setup](docs/codex/setup.md) - [Codex live-test handoff](docs/codex/live-test-handoff.md) - [Codex acceptance walkthrough](docs/walkthroughs/codex-acceptance-demo.md) - [Acceptance analytics](docs/analytics/acceptance-analytics.md) - [Evidence dashboard](docs/analytics/evidence-dashboard.md) - [Event ledger](docs/analytics/event-ledger.md) - [Golden-case harness](docs/evals/golden-case-harness.md) - [Model registry](docs/configuration/model-registry.md) - [Dogfooding guide](docs/dogfooding/phase5-dogfooding.md) - [Goose demo walkthrough](docs/walkthroughs/goose-acceptance-demo.md) - recording-ready 3-5 minute public demo runbook - [Policy packs](docs/policy-packs/) - [PR gate](docs/github/pr-gate.md) - [Launch issues](docs/github/launch-issues.md) - [Repository topics](docs/github/repository-topics.md) - [Create launch issues](docs/github/create-launch-issues.md) - [Publishing guide](docs/publishing/pypi.md) - [Gemini fixture proof](docs/proof/gemini-fixture-accepted-run.md) - [Codex fixture proof](docs/proof/codex-fixture-accepted-run.md) Recipes: - [Engineering acceptance](recipes/workbench-engineering-acceptance.yaml) - [MCP tool smoke](recipes/workbench-mcp-tool-smoke.yaml) - [Docs-only acceptance](recipes/workbench-docs-only-acceptance.yaml) - [Python package maintenance](recipes/workbench-python-package-maintenance.yaml) - [Test-fix acceptance](recipes/workbench-test-fix-acceptance.yaml) Sample evidence: - [Accepted tiny Python fix](examples/sample-runs/accepted-tiny-python-fix) - [Accepted Codex tiny Python fix](examples/sample-runs/accepted-codex-tiny-python-fix) - [Accepted docs-only smoke](examples/sample-runs/accepted-docs-only-smoke) - [Needs-review test fix](examples/sample-runs/needs-review-test-fix) ## License Apache-2.0. See [LICENSE](LICENSE). MIT-origin attribution for the consolidated Prove It code is retained in [NOTICE](NOTICE).