Agentra LabsAgentra Labs DocsPublic Documentation

AgenticPlanning

Installation

AgenticPlanning can be installed through four methods. Choose the one that fits your workflow.

AgenticPlanning can be installed through four methods. Choose the one that fits your workflow.

Requires Rust 1.75+:

cargo install agentic-planning-cli

For the standalone MCP server binary:

cargo install agentic-planning-mcp

This builds and installs the aplan binary to ~/.cargo/bin/. The binary includes the CLI, MCP server (aplan serve), and daemon mode. The standalone agentic-planning-mcp binary is useful for deployments that only need the MCP server.

Pre-built Binary

Platform-detected installer:

curl -fsSL https://agentralabs.tech/install/planning | bash

The installer detects your platform (macOS arm64/x86_64, Linux x86_64), downloads the binary, places it in ~/.local/bin/, and adds it to PATH if needed.

Verify:

aplan version

npm Package (WASM)

For Node.js or browser usage:

npm install @agenticamem/planning
const { PlanningEngine } = require('@agenticamem/planning');

const engine = PlanningEngine.inMemory();
const goalId = engine.createGoal({
  title: "Ship feature",
  intention: "Deliver value to users",
  priority: "high"
});

pip Package (Python FFI)

For Python usage via ctypes:

pip install agentic-planning
from agentic_planning import PlanningEngine

engine = PlanningEngine.in_memory()
goal_id = engine.create_goal(
    title="Ship feature",
    intention="Deliver value to users",
    priority="high"
)

The pip package bundles the native shared library and provides Python bindings over the 27 FFI functions.

From Source

git clone https://github.com/anthropics/agentic-planning.git
cd agentic-planning
cargo build --release

Binaries are placed in target/release/:

BinaryPurpose
aplanCLI + MCP server + daemon
agentic-planning-mcpStandalone MCP server

MCP Server Setup

To use AgenticPlanning with Claude Code, add the MCP server configuration to your settings:

Claude Code (~/.claude/settings.json):

{
  "mcpServers": {
    "agentic-planning": {
      "command": "aplan",
      "args": ["serve"]
    }
  }
}

With a persistent state file:

{
  "mcpServers": {
    "agentic-planning": {
      "command": "aplan",
      "args": ["serve", "--file", "/path/to/project.aplan"]
    }
  }
}

The MCP server communicates over stdio using Content-Length framed JSON-RPC 2.0 — the standard MCP transport protocol.

Verification

After installation, confirm everything works:

# Check version
aplan version

# Check engine status
aplan status

# Create a test goal
aplan goal create "Test goal" --intention "Verify installation"

# List goals
aplan goal list

Upgrading

  • Cargo: cargo install agentic-planning-cli (rebuilds latest)
  • Binary: Re-run the curl installer
  • npm: npm update @agenticamem/planning
  • pip: pip install --upgrade agentic-planning

The .aplan file format is version-safe across minor releases. The header version field ensures older files are loaded correctly by newer engines.