Fleet Prime Agent

Fleet Prime is a local, multi-project workspace for coding and research with AI. It combines a focused web interface with the stock Prime Agent runtime, keeping sessions, workspace navigation, live tool activity, and managed execution together on your machine.

fleet-prime-0.5.8.mp4

Quick start

Install the package

Fleet supports macOS and Linux; Windows is not supported. You need Node.js 22.12.0 or later, and Python 3.10 or later for the managed IPython kernel.

npm install --global @qredence/fleet

Launch the web workspace

Run Fleet from the project directory you want the agent to work in:

cd /path/to/your/project
fleet-agent

Open the loopback URL printed by the launcher. On first use, add a provider in Settings → Providers (or run /login) and choose a model in the composer.

fleet-prime is an alias for the web launcher.

Terminal mode

Run the terminal interface with:

fleet-agent agent

An existing prime-agent installation remains untouched. Fleet uses the same upstream settings, kernel environment, and logs.

Install from source

Use the source installer when you want to develop Fleet or run the repository checkout directly:

git clone https://github.com/Qredence/fleet-prime-agent.git
cd fleet-prime-agent
./fleet-prime.sh install

The installer installs workspace dependencies, builds the web runtime, and places a fleet-agent launcher in a user-writable bin directory. If that directory is not already on your PATH, the installer prints the path to add. Source installation also requires Git, Node.js, and npm.

What Fleet provides

• A local workspace for multiple projects and persistent sessions.
• Streaming assistant responses and dedicated activity for shell commands, file edits, plans, questions, and subagents.
• A managed IPython environment for interactive analysis and file operations.
• Attachments, generated artifacts, workspace browsing, and session history.
• Plan, refinement, and provider configuration flows through the web interface.

Develop

Install the repository workspace with pnpm:

pnpm install

Start the web application in development mode:

pnpm run dev:web

Run the repository checks and tests before submitting a change:

pnpm run check
pnpm run test:web

Use pnpm for the repository workspace. Do not run npm install or npm ci at the repository root; those commands create an unsupported dependency layout. Contributors should read CONTRIBUTING.md and AGENTS.md before opening a pull request.

Architecture

Fleet owns the product and adapter layers; Prime Agent remains the external execution engine. The browser communicates with Fleet through typed HTTP, NDJSON, and SSE contracts, while runtime access stays in web/server.

The pinned runtime release and checksum are maintained in PRIME_AGENT_RUNTIME.json.

Security

Fleet runs locally and the web launcher accepts loopback connections only. It can execute model-generated Python and project commands with your user permissions; it is not a security sandbox. Use trusted repositories, instructions, skills, and extensions, and review changes before accepting them. See SECURITY.md for vulnerability reporting.

Documentation and support

• Architecture — system ownership and data flow.
• Adapter contract — browser/server compatibility, replay, and privacy guarantees.
• Contributing — development and pull request process.
• Release guide — Changesets, publishing, and rollback.
• Support — questions, bugs, and community channels.
• Prime Agent documentation — upstream engine documentation.

License

Fleet Prime Agent is released under the MIT License.

Fleet is powered by Prime Agent, whose lineage includes pi-mono by Mario Zechner.