Build with Hilla.

Hilla is an all-in-one project canvas. Plan, build, track, and ship on the same surface. You can interact with Hilla through the web app, CLI, or any AI tool that supports the Model Context Protocol.

This documentation covers the developer-facing tools: the CLI for terminal workflows, the MCP server for AI-native integrations, and the REST API for programmatic access.

Install the Hilla CLI globally via npm. Requires Node.js 18 or later.

$ npm install -g @hilla/cli

$ hilla --version
hilla/1.0.0

Authenticate with your Hilla account to connect the CLI to your projects.

$ hilla auth login
→ Opening browser for authentication...
✓ Authenticated as you@example.com

Generate your first project board in seconds. Describe what you want to build and Hilla handles the rest — market research, architecture, sprint plans, and task breakdown.

$ hilla plan "Build a SaaS for restaurant analytics"

→ Researching market landscape...
→ Generating architecture...
→ Creating sprint plan...

✓ 3 sprints  ·  47 tasks  ·  12 dependencies

→ Board ready at hilla.ai/b/rest-analytics

Open the board URL in your browser to see the full spatial canvas, or continue working from the CLI.

The Hilla CLI lets you create and manage projects, generate plans, execute tasks, and export boards — all from your terminal. Designed for developers who prefer keyboard-driven workflows.

Every command supports --json output for scripting and automation. Use hilla help <command> for detailed usage.

Chain commands with standard Unix pipes. For example, export a board and pipe it to another tool:

$ hilla export --json | jq '.tasks[] | select(.sprint == 1)'

The CLI stores configuration in ~/.hilla/config.json. You can edit it directly or use hilla config to update values.

The Model Context Protocol is an open standard that lets AI tools connect to external services. Hilla runs an MCP server so your AI coding assistant can create, read, and manage cards on your board directly — without switching context.

When you connect Hilla via MCP, your AI tool gains access to your project boards, tasks, and sprint plans. It can suggest tasks, update progress, and even execute work on your behalf.

Start the MCP server locally. It listens on stdio by default, which is what most AI clients expect.

$ hilla mcp start
→ MCP server running (stdio)
→ Waiting for client connection...

Add Hilla to your AI tool's MCP configuration. Here's the standard entry:

{
  "mcpServers": {
    "hilla": {
      "command": "hilla",
      "args": ["mcp", "start"],
      "env": {
        "HILLA_API_KEY": "your-api-key"
      }
    }
  }
}

The MCP server exposes these tools to connected AI clients. Each tool maps to a Hilla API operation.

Any AI tool that implements the Model Context Protocol can connect to Hilla. These clients have been tested and confirmed working:

All API requests require a bearer token. Generate one from your account settings or via the CLI.

$ hilla auth token
→ Your API token: hilla_sk_...

# Use in requests:
$ curl -H "Authorization: Bearer hilla_sk_..." \
    https://api.hilla.ai/v1/projects

Tokens don't expire but can be revoked from your account settings at any time.

Create and manage project boards programmatically.

$ curl -X POST https://api.hilla.ai/v1/projects \
    -H "Authorization: Bearer hilla_sk_..." \
    -H "Content-Type: application/json" \
    -d '{"prompt": "Build a SaaS for restaurant analytics"}'

{
  "id": "proj_rest_analytics",
  "name": "Restaurant Analytics SaaS",
  "tasks": 47,
  "sprints": 3,
  "url": "https://hilla.ai/b/rest-analytics"
}

Read and manage individual task cards on a project board.

$ curl https://api.hilla.ai/v1/projects/proj_rest_analytics/tasks?sprint=1 \
    -H "Authorization: Bearer hilla_sk_..."

{
  "tasks": [
    {
      "id": "task_001",
      "title": "Auth system (OAuth + magic link)",
      "sprint": 1,
      "status": "todo",
      "dependencies": []
    },
    ...
  ]
}