Skip to main content
v0.10.3 — Apache-2.0 · Open Source

The AI Work Harness
That Keeps Getting Better

Capture your best AI behavior, refine it as a team, and deploy it to every repo automatically.

~/code/my-org
$agentboot install# creates hub, builds & syncs
✓ Scaffolded personas/  ✓ Built 4 personas × 8 platforms
✓ Synced → 6 repos across 2 teams
$cd acme-widget-service && claude
Claude Code started
✦/ab How do I import all of our company's awesome prompts,
skills, and personas all in one fell swoop?
⏺That's an import — pulling your existing agentic content into the hub. Let me scan first so you know exactly what would be imported before anything changes.
⏺agentboot — agentboot_scan_for_import (MCP)
⎿  { "highConfidence": 12, "needsReview": 3, ... }
8
Output Platforms
4
Core Personas
0
Tests Passing
Apache-2.0
Licensed
0
TypeScript Errors

How It Works

Three steps from authoring to deployed AI behavior across your entire organization.

Step 01

Author

Write traits and personas in markdown

Define AI behavior as source code. Traits are reusable behavioral building blocks — critical thinking, source citation, structured output. Compose them into personas with a single config file.

# core/traits/critical-thinking.md --- name: critical-thinking weight: HIGH --- Examine claims before accepting them. Flag assumptions. Request evidence.
Step 02

Compile

One command, 8 platform-native formats

`agentboot build` resolves trait references, inlines weights, and emits platform-native output simultaneously. Claude Code gets `.claude/agents/`. Copilot gets `.github/copilot-instructions.md`. Cursor gets `.cursor/rules/`. All from one source.

$ agentboot build dist/claude/ → .claude/agents/ dist/copilot/ → .github/ dist/cursor/ → .cursor/rules/ dist/gemini/ → GEMINI.md dist/agents/ → AGENTS.md dist/windsurf/ → .windsurfrules +2 more platforms
Step 03

Deploy

Sync to every repo in one step

`agentboot sync` reads your repos list, applies scope merging (org → group → team), and writes platform-native files to every target repo. Manifests track file hashes so re-syncs are idempotent.

$ agentboot sync Syncing core personas... ✓ github/api-service (claude, copilot) ✓ github/frontend-app (claude, cursor) ✓ github/data-pipeline (gemini, agents) ✓ github/mobile-ios (copilot, jetbrains) 12 repos updated · 0 conflicts

8 Platforms. One Source of Truth.

Every platform gets native output — no wrapper scripts, no adapter layers. AgentBoot speaks each platform's dialect fluently.

Claude Code
.claude/
GitHub Copilot
.github/
Cursor
.cursor/rules/
Windsurf
.windsurfrules
JetBrains AI
.junie/
Gemini Code Assist
GEMINI.md
AGENTS.md
Universal standard
agentskills.io
SKILL.md

Built for Two Audiences

Engineers get a work harness that doesn't get in their way. Organizations get consistent AI behavior across every repo.

Your work harness. Your personal setup stays yours.

For Engineers

Battle-tested personas for code review, security analysis, test generation, and data modeling. AgentBoot is what your org runs — your personal AI setup is untouched.

  • ✓Code Reviewer — finds real bugs, not style nits
  • ✓Security Reviewer — adversarial threat modeling
  • ✓Test Generator — coverage-aware, pattern-respecting
  • ✓Test Data Expert — realistic synthetic fixtures
See how it works →
Centralized governance, local autonomy

For Organizations

Define AI behavior at the org level. Distribute to every repo. Enforce compliance centrally while teams customize locally. Four-level scope hierarchy: org → group → team → repo.

  • ✓One-command deployment to all repos
  • ✓Scope hierarchy — team overrides group overrides org
  • ✓Manifest tracking — idempotent re-syncs
  • ✓Compliance hooks — 3-layer defense-in-depth
Architecture Overview →

Quick Start

Up and running in under 5 minutes. No AI provider account required for core features.

~/code/my-org
$npm install -g agentboot# install globally
$agentboot install# creates hub, builds & syncs — done
✓ Scaffolded personas/  ✓ Built 4 personas × 8 platforms  ✓ Synced → all repos
$cd acme-widget-service && claude
✦/ab How do I import all of our company's awesome prompts, skills, and personas all in one fell swoop?
⏺That's an import — pulling your existing agentic content into the hub. Let me scan first so you know exactly what would be imported before anything changes.
⏺agentboot — agentboot_scan_for_import (MCP)(paths: ["~/code/api-service", "~/code/frontend", ...])
⎿  { "highConfidence": 12, "needsReview": 3, "skipped": 1 }

Start in 5 minutes.

No AI provider account needed for core features. Apache-2.0 licensed. Self-hosted. Your prompts never leave your machine.

Read the Getting Started Guide →