Agent Dashboard MCP Server

<p align="center"> <a href="LICENSE"><img src="https://img.shields.io/badge/license-Apache--2.0-blue" alt="License: Apache-2.0"></a> <a href="https://github.com/trustabl/agent-dashboard/releases"><img src="https://img.shields.io/github/v/release/trustabl/agent-dashboard" alt="Latest release"></a> <a href="https://github.com/trustabl/agent-dashboard/stargazers"><img src="https://img.shields.io/github/stars/trustabl/agent-dashboard?style=flat" alt="GitHub stars"></a> <a href="https://github.com/trustabl/agent-dashboard/actions/workflows/ci.yml"><img src="https://img.shields.io/github/actions/workflow/status/trustabl/agent-dashboard/ci.yml?branch=main&amp;label=CI" alt="CI"></a> <a href="package.json"><img src="https://img.shields.io/badge/node-%E2%89%A522.13-339933" alt="Node.js 22.13 or newer"></a> <a href="package.json"><img src="https://img.shields.io/github/package-json/dependency-version/trustabl/agent-dashboard/next?label=Next.js" alt="Next.js version"></a> <br> <a href="https://github.com/trustabl/agent-reliability-analyzer"><img src="https://img.shields.io/github/v/release/trustabl/agent-reliability-analyzer?label=analyzer%20%26%20forge" alt="Latest Analyzer and Forge release"></a> <a href="https://github.com/trustabl/agent-network-access-probe"><img src="https://img.shields.io/github/v/release/trustabl/agent-network-access-probe?label=probe" alt="Latest Probe release"></a> <a href="#first-run--5-minutes"><img src="https://img.shields.io/badge/platforms-macOS%20%C2%B7%20Linux%20%C2%B7%20WSL2-blue" alt="Platforms: macOS, Linux, WSL2"></a> <a href="README.md"><img src="https://img.shields.io/badge/runs%20on-127.0.0.1%20only-blue" alt="Runs on 127.0.0.1 only"></a> </p>

Agent Dashboard

Agent dashboard runs reliability scans, skills and network access probes on your own repos from one local, single-user web UI. It wraps Trustabl's open-source Analyzer, Forge and Probe, so you don't have to install any of them. Policy and telemetry via OpenTelemetry are coming soon.

Page	Wraps	Commands
Overview	all of the below	—
Forge	agent-reliability-analyzer	trustabl forge, trustabl forge check
Analyzer	agent-reliability-analyzer	trustabl scan
Probe	agent-network-access-probe	autofix fix

• Runs on 127.0.0.1 only. There is no login and no user or org model.
• Pinned CLIs: trustabl 0.1.16 and autofix 0.5.0. make run downloads them from their GitHub releases and checks their SHA-256 (tools.lock.json). Newer releases can be installed from the Forge, Analyzer and Probe pages.
• Your source code stays on your machine. Probe sends tool source only to the LLM endpoint you configure.

First run (≈ 5 minutes)

Requirements: macOS, Linux or WSL2, with git, curl, tar and make. You need Node.js 22.13 or newer; if you don't have it, make run installs a private copy. To run Probe you also need Docker (or OpenShell) and an OpenAI-compatible LLM endpoint, for example Ollama.

git clone https://github.com/trustabl/agent-dashboard.git
cd agent-dashboard
make run            # or: ./scripts/run.sh

Open http://127.0.0.1:4317 and click Generate skill, Scan repo or Run Probe. Pick New local path — agent not built, enter a folder (for example ~/src/my-agent) and click Start. The output streams live. The run keeps going if you close the tab, and you can cancel it at any time. Every run has Copy CLI, which copies the exact command to run in a terminal.

For Probe with a local model:

ollama serve & ollama pull qwen2.5-coder:32b
# in the dialog: --model qwen2.5-coder:32b, --llm-url http://localhost:11434/v1

Cloud models read their key from OPENAI_API_KEY, ANTHROPIC_API_KEY or LLM_API_KEY in the environment you start the dashboard from. The dashboard never stores keys.

Where state lives, and how to wipe it

All state is under ~/.agent-dashboard (override with AGENT_DASHBOARD_HOME):

Path	What
console.db	Repos, run history, findings (SQLite)
runs/<id>/	Each run's command, output and artifacts (scan.json, trustabl.sarif, SKILL.md, probe.*.json, network_policies.draft.yaml)
bin/	CLIs, one folder per installed version
cache/	The CLIs' rule and Probe result caches (Linux/WSL)

make wipe           # asks first; WIPE_YES=1 make wipe to skip the prompt

This deletes everything above. On macOS it also deletes the CLIs' caches in ~/Library/Caches/trustabl*. Your repos are not touched, and neither is any SKILL.md you chose to write into them.

Development

make dev            # hot reload on 127.0.0.1:4317
make check          # eslint + tsc + vitest
make e2e            # Playwright against tests/fixtures/repos (Probe cases need Docker)
make parity         # artifacts from a dashboard run == the same CLI run (timestamps masked)
make secrets        # gitleaks, as in CI

See docs/ARCHITECTURE.md for the architecture (components, API routes, run engine, data layer, conventions). Contributions: CONTRIBUTING.md.

License

Apache-2.0. See LICENSE.