Skip to content

Welcome!

PlanExe turns a single plain-English goal into a ~40-page strategic plan in ~15 minutes using local or cloud LLMs. It’s an accelerator for first drafts — not a replacement for human refinement. PlanExe removes most of the labor for the planning scaffold; the final 10–30% that makes a plan credible and defensible remains human work.

Try it first, then decide if you want to run it locally.

Start here (pick your path)

Use the short decision guide: Start here

Core guides

What you get

PlanExe generates a single HTML report (a self-contained artifact you can open in a browser). See the sample report here: Minecraft Escape sample report

2-minute tour

Open the sample report and do this:

  1. Read Executive Summary to see the top-level deliverables, budget, risks, and next steps.
  2. Jump to Gantt Interactive to see how the goal gets broken down into many concrete tasks.
  3. Open Premortem to see what could go wrong and what to do about it.

Get help

If you run into issues, join the PlanExe Discord — the community and maintainers are there to help. When you ask, include what you tried, your setup (OS, Docker vs local, model/LLM), and any error output so others can help you quickly.

Join the PlanExe Discord →

Model Context Protocol (MCP)

PlanExe provides an MCP server for AI agents.

The tool workflow is:

  1. Call example_plans to preview the output. This step is optional.
  2. Call example_prompts.
  3. Call model_profiles to choose a model profile. This step is optional.
  4. Draft and approve a detailed prompt.
  5. Call plan_create.
  6. Poll plan_status about every five minutes until the plan is complete.
  7. Call plan_retry if generation fails and should restart.
  8. Download the report or source bundle with plan_file_info.

Each plan_create call returns a new plan_id. Clients are responsible for tracking plans they create in parallel.

Connect to the hosted MCP server

You need:

  • An account at home.planexe.org
  • Enough account credit to generate a plan
  • A PlanExe API key beginning with pex_

Add this configuration to an MCP-compatible client:

{
  "mcpServers": {
    "planexe": {
      "url": "https://mcp.planexe.org/mcp",
      "headers": {
        "X-API-Key": "pex_your_api_key_here"
      }
    }
  }
}

PlanExe has setup guides for:

See MCP setup, tool details, and the PlanExe MCP interface for the full reference.

Run the MCP server locally

You need Docker and an LLM provider configuration. Start the stack:

docker compose up --build

Confirm that the web interface can create a plan, then connect the MCP client to:

http://localhost:8001/mcp

Authentication is disabled by default in the local Docker configuration.

Run the pipeline from the command line

Run these commands from the root of a PlanExe checkout:

# For a substantial project, put the full brief in a file
./planexe create_plan \
    --plan-file regional_clean_water_brief.txt \
    --output-dir ./planexe-outputs/1984-12-31/RegionalCleanWater_v1

# A shorter prompt can also be supplied directly
./planexe create_plan \
    --plan-text "Plan a phased ERP migration across five countries without interrupting manufacturing" \
    --output-dir ./planexe-outputs/1984-12-31/ManufacturingERP_v1

The command creates the output directory and writes:

File Contents
start_time.json Start time in UTC
plan.txt The submitted project brief

Use PlanExe from an AI agent

An AI agent can create a plan through the MCP server and use the report or source bundle in later work.

  • Service metadata: mcp.planexe.org/llms.txt
  • MCP endpoint: https://mcp.planexe.org/mcp
  • Prompt format: call example_prompts, then draft 300–800 words of flowing prose covering the objective, constraints, timeline, stakeholders, resources, and success criteria
  • Agent workflow: Autonomous agent guide

The downloadable source bundle includes Markdown, JSON, and CSV artifacts such as the work breakdown structure and pre-project assessment.

Run PlanExe locally with Docker

Docker Compose runs the web interface, plan worker, and supporting services.

1. Clone the repository

git clone https://github.com/PlanExeOrg/PlanExe.git
cd PlanExe

2. Configure a model provider

Copy .env.docker-example to .env, then add the required credentials for the provider you selected. For example, an OpenRouter setup uses OPENROUTER_API_KEY.

The containers mount .env and llm_config/. To use Ollama on the host, select the relevant Ollama profile and make sure it is listening on http://host.docker.internal:11434.

3. Start PlanExe

docker compose up worker_plan frontend_multi_user

The web interface becomes available at http://localhost:5001 after its dependencies pass their health checks. Generated plans are written to run/.

To watch the worker:

docker compose logs -f worker_plan

Stop the services with Ctrl+C.

4. Rebuild after dependency changes

docker compose build --no-cache worker_plan frontend_multi_user

See Docker setup for ports, configuration, and troubleshooting.

Model configuration

Cloud models are generally simpler to configure. Local models give you more control over the model and where project data is processed.

Next steps