MCP et appels d’outils

human

gethuman-sh/human

An open-source AI Software Factory that integrates tickets, docs, designs, and analytics into a secure pipeline, outputting shipped code via Claude-driven agents.

Ouvrir le dépôt officiel ↗ Page d'accueil du projet ↗
★ 62Étoiles
⑂ 6Forks
3Problèmes ouverts
GoLangue
MITLicence
Q@project.QualityScoreScore éditorial

Captures d'écran du projet

Capture d’écran de human Capture d’écran de human Capture d’écran de human Capture d’écran de human Capture d’écran de human Capture d’écran de human

Vue d’ensemble

human is an AI development rig designed to provide teams with a pre-configured, secure pipeline from idea to code review. It integrates with existing issue trackers, documentation, design, and analytics platforms, wiring them together around a coding agent. The system emphasizes context engineering, token efficiency, and strict security boundaries, ensuring the AI operates on real requirements without exposing sensitive credentials. It features a daemon-client architecture for isolated devcontainers, a desktop workflow board, and a suite of lifecycle skills for autonomous development.

Fonctionnalités clés

  • Secure by default: Outbound firewall, secret-redacting filesystem, vault-resolved credentials, and isolated devcontainers.
  • Token efficiency: Structural code navigation and signal-extracted data reduce token usage by up to 95% compared to raw APIs.
  • Context engineering: Connectors for trackers, Notion, Figma, and Amplitude with cross-source search and codebase indexing.
  • Daemon-client architecture: Host daemon holds credentials and proxies requests; container client forwards commands securely.
  • Lifecycle skills: Pre-built Claude Code skills for ideate, sprint, plan, execute, review, autofix, security, and gardening.
  • Desktop workflow board: Interactive drag-and-drop GUI to trigger pipeline stages and monitor agent progress.
  • Audit trail: Structured, queryable record of every agent action against trackers, with durable on-disk logs.

Prérequis, installation et démarrage rapide

Via curl: `curl -sSfL gethuman.sh/install.sh | bash`
Via Homebrew: `brew install --cask gethuman-sh/tap/human`
Via mise: `mise use -g github:gethuman-sh/human`
Via Go: `go install github.com/gethuman-sh/human@latest`
As a devcontainer Feature: Add `{ "features": { "ghcr.io/gethuman-sh/treehouse/human:1": {} } }` to `devcontainer.`.

Utilisation

Run `human init` to start the interactive setup wizard, which configures services and generates `devcontainer.`. Set the required API tokens printed by the wizard. Start the daemon with `human daemon start`. Launch the devcontainer with `devcontainer up --workspace-folder .`. Use CLI commands like `human get <KEY>`, `human list`, `human search <query>`, or `human pr create`. For Claude Code integration, run `human install --agent claude` to install skills, then use skills like `/human-sprint` or `/human-autofix`.

Compatibilité des modèles et cas d’usage

The repository explicitly mentions integration with Claude Code (Anthropic's Claude) for its AI developer agents and skills.

Licence et notes sur les risques

MIT License.

Editorial verification 2026-08-02: repository URL, owner, description, license and repository statistics were reviewed. License metadata: MIT. README was fetched for the channel draft; re-check repository dependencies, releases and model terms before production use.

Publication et maintenance

Not stated in the repository metadata

MemPalace

mempalace/mempalace

★ 58KPython

QwenPaw

agentscope-ai/qwenpaw

★ 33,9KPython

Composio

composiohq/composio

★ 29,5KTypeScript