Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.getpioneer.dev/llms.txt

Use this file to discover all available pages before exploring further.

Pioneer is a local-first AI workspace for running an assistant on your own machine or on infrastructure you control. It combines a persistent gateway, a native desktop app, a JSON-RPC protocol, provider adapters, durable threads, task automation, MCP servers, skills, and real local tools. The gateway is the core of Pioneer. It owns state, configuration, storage, model access, task execution, tool execution, MCP runtime, skills, and thread history. Secret values such as provider API keys, MCP env/header secrets, and superuser JWT signing material live in the gateway keystore. The desktop app is the primary client for connecting to one or more gateways, whether the gateway is running on the same computer or on a remote server.
Pioneer is in an extremely early stage of development. Expect rough edges, breaking changes, and incomplete flows. Use it carefully, and test it in a safe environment before trusting it with important work or exposing a gateway outside your machine.
Tool runs are not sandboxed yet. Tools currently execute as the OS user running the gateway.

Rust Native

Pioneer is built in Rust all the way through the product: gateway, CLI, desktop app, protocol, tools, tasks, MCP, skills, and provider integrations. That keeps the core memory-safe, fast, and small. The desktop app is native GPUI, not Electron or a web app wrapped in a window.

Gateway And Desktop

Pioneer is split into three user-facing pieces. Gateway is the main runtime and control plane. It runs as a service, stores data in gateway.db, stores secret values in keystore.db, talks to model providers, executes tools, manages MCP servers and skills, schedules tasks, and exposes the JSON-RPC WebSocket API. Desktop app is the primary native client. It connects to gateways, starts and manages a local gateway when needed, and gives you the UI for conversations, provider setup, MCP, skills, settings, and thread history. Protocol clients are any clients built on top of the Pioneer JSON-RPC protocol. The desktop app is the main client today, and native mobile apps for iOS and Android are planned next. For a single-machine setup, install the desktop app and let it start the local gateway for you. On macOS, that means downloading the .dmg, moving Pioneer to Applications, launching it, and pressing Start local gateway when prompted. For a multi-environment setup, install gateways wherever the work should live: a laptop, a workstation, a home server, or a remote machine. Then connect to each gateway from the same desktop app. You can keep separate gateways for work, study, home, experiments, or clients without mixing their state, settings, tools, and histories.

Install Pioneer

Set up the desktop app, a local gateway, or a remote gateway.

Quick Start

Connect a provider and start your first thread.

Core Concepts

Understand gateways, the default workspace, threads, tools, skills, MCP, and tasks.

Providers

Connect Anthropic, OpenAI, OpenRouter, Ollama, Google, Azure, Bedrock, and more.

Multi-Agent Workflows

Pioneer is designed for more than one long-running chat thread. The gateway can coordinate subagents automatically as part of the task system: a parent task can create an agent spec, start a child thread, run that subagent with its own model and instructions, and link the result back to the parent task. Each subagent can be scoped independently:
ScopeWhat it controls
Role and identityGive the agent a role and nickname so its work is understandable in task history.
Model choiceChoose the model and provider per agent instead of forcing every subtask through the same model.
Context policyInherit parent context, pass only recent turns, use a summary, start empty, or provide custom context.
Tool policyAllow or deny tools, choose read-only or write-capable modes, restrict paths, and control network access.
Result contractAsk for text, Markdown, JSON, or artifacts with required outputs.
Depth and lineageTrack nested agent work through child threads and task lineage so delegated work can be audited and recovered.
This makes the gateway useful as a coordinator. One agent can break work into subtasks, specialized subagents can handle pieces in isolated threads, and the desktop app can still show the full task tree from one place. Users shape this with goals and policies, while Pioneer handles the orchestration.

Where To Start

Start with Installation, then follow Quick Start. After that, use Core Concepts when a term in the UI is unclear, Providers when you want to connect models, Skills when you want reusable capabilities, MCP Servers when you want external tools, and Tasks when you want repeatable or scheduled work.