AutoProject

AutoProject is your ultimate Bring Your Own Keys (BYOK) tool to automate product and project management with AI-powered workflows. It combines a SvelteKit 2 + Svelte 5 frontend for human interaction with a Bun/Hono backend + MCP server that drives an AI-agent workflow engine, all deployable as a single Docker container.

---

Architecture

autoproject/          SvelteKit 2 + Svelte 5 UI (port 3000)
  └── src/            AI-powered PRD, user stories, Linear integration, project view

backend/              Bun/Hono HTTP API + MCP server + workflow engine (port 3001)
  ├── src/
  │   ├── app.ts              Hono app (37 HTTP endpoints)
  │   ├── mcp/                MCP server (32 tools, 10 resources)
  │   ├── llm/                LLM client (OpenAI-compatible abstraction)
  │   ├── shared/             ApiError, response helpers, validation
  │   └── workflow-engine/    SQLite state machine, prompts, quality gates
  ├── dist/                   Compiled backend
  └── package.json

Dockerfile                Multi-stage build — both layers in one image
docker-entrypoint.sh      Starts backend then serves SvelteKit on port 3000

The frontend and backend are co-deployed in a single container. The SvelteKit app on port 3000 communicates with the backend API on port 3001. AI agents use the MCP server (stdio) as their primary integration point; HTTP endpoints mirror the same operations for non-agent automation.

---

Key Features

• AI PRD Creation: Feed your requirements and receive tailor-made PRDs.
• AI User Story Generation: Convert PRDs into smart, structured user stories.
• Project View: Manage your user stories, copy their details for AI prompting.
• Tool Integration: Instantly push stories into Linear.
• AI-Agent Workflow Engine: Requirements become incremental steps stored in SQLite; agents orchestrate the full lifecycle (scenarios, design, tests, gates).
• Multiple Inference Support: Choose from Gemini, Groq, OpenAI, Anthropic, LM Studio, or any OpenAI-compatible endpoint.
• Preset Modes: Specialized configurations for different use cases.

---

Running with Docker (Recommended)

Using Docker Compose

Ensure Docker Desktop is installed and the engine is running.

Prepare your .env file in autoproject/:

# PROJECT TOOLS API TOKENS
SECRET_LINEAR_API_KEY="Your linear api key"
SECRET_JIRA_API_TOKEN="Your jira api token"
SECRET_ASANA_API_TOKEN="Your asana api token"
SECRET_PLANE_API_TOKEN="Your plane api token"

# AI API TOKENS
SECRET_GEMINI_API_KEY="Your gemini api key"
SECRET_GROQ_API_KEY="Your groq api key"
SECRET_OPENAI_API_KEY="Your openai api key"
SECRET_ANTHROPIC_API_KEY="Your anthropic api key"

# Local inference
INFERENCE_SERVER_URL=http://127.0.0.1:1234
INFERENCE_SERVER_URL_DOCKER=http://host.docker.internal:1234

# LLM provider for the backend workflow engine (optional)
LLM_BASE_URL=http://127.0.0.1:1234/v1
LLM_MODEL=qwen/qwen3-1.7b

docker compose up -d --build

Access the app at http://localhost:3000.

AutoProject stores SQLite state under /data/.autoproject/autoproject.db. The named volume keeps this data safe across restarts.

docker compose stop
docker compose down      # keep volume
docker compose down -v    # delete all data

---

Development Setup

Prerequisites

• Bun (latest version)
• Docker Desktop (optional, for Docker testing)
• API keys for your chosen AI providers

Frontend (SvelteKit)

cd autoproject
bun install
cp .env.example .env        # Edit with your real API keys
bun run dev                  # Dev server at http://localhost:5173
bun run build                # Production build
bun run preview              # Preview at http://localhost:4173

Backend (Hono API + MCP)

cd backend
bun install
bun run dev                  # HTTP API on http://127.0.0.1:3001
bun run mcp                  # MCP server on stdio

See backend/README.md for full MCP tool and HTTP endpoint reference.

---

Environment Variables

Variable	Required	Description
SECRET_LINEAR_API_KEY	✅	Linear API access key
SECRET_GEMINI_API_KEY	✅	Gemini API access key
SECRET_GROQ_API_KEY	✅	Groq API access key
SECRET_OPENAI_API_KEY		OpenAI API key (for LLM or AI SDK)
SECRET_ANTHROPIC_API_KEY		Anthropic API key
SECRET_JIRA_API_TOKEN		Jira API token
SECRET_ASANA_API_TOKEN		Asana API token
SECRET_PLANE_API_TOKEN		Plane API token
INFERENCE_SERVER_URL		Local inference URL (default http://127.0.0.1:1234)
INFERENCE_SERVER_URL_DOCKER		Docker host inference URL
LLM_BASE_URL_DOCKER		Docker host LLM URL override
PORT		Web server port (default 3000)
HOST		Web server host (default 0.0.0.0)
BACKEND_PORT		Backend API port (default 3001)
BACKEND_HOST		Backend API host (default 127.0.0.1)
AUTOPROJECT_BACKEND_URL		Backend URL for web server (Docker auto-set)
DOCKER_ENV		Set to true inside Docker container
LLM_BASE_URL		OpenAI-compatible API base for backend
LLM_API_KEY		API key for the LLM provider
LLM_MODEL		Default LLM model for backend
AUTOPROJECT_ROOT		Workspace root (default /data in Docker)

No secrets are included in the public Docker image. All sensitive information must be provided at runtime via .env.

---

Core Workflow (AI-Agent MCP)

requirements → prompts_execute (step-generation) → steps_import
  └→ for each step: test-scenarios → design → tests → implement → gates → complete

See AUTOPROJECT.md for the full methodology and agent work loop.

---

Testing

# Backend tests
cd backend
bun test                     # All tests (600s timeout)
bun run test:unit            # Unit tests only (120s timeout)
bun run test:integration     # Service-level integration tests
bun run test:mcp             # MCP tool/resource tests
bun run test:live            # All live tests (workflow + LLM)
bun run test:live:workflow   # Live HTTP workflow E2E
bun run test:live:llm        # Live LLM integration (needs LLM provider)

See TESTING.md for the full testing checklist and backend/TESTING.md for backend-specific test details.

---

Contributing

Contributions are welcome! Feel free to fork, clone, and submit PRs.

---

License

Licensed under the MIT License.