BrainPilot v0.1.0 — First Open-Source Release

BrainPilot v0.1.0 is our first open-source release. BrainPilot is a human-in-the-loop agentic research system for brain science that coordinates specialist agents, domain knowledge, scientific skills, and tools while preserving an inspectable research process through Graph of Trace.

Highlights

• Built for brain science research — supports literature review, hypothesis refinement, experiment design, data analysis, writing, and audit.
• PI Agent + specialist agents — a coordinating PI works with a librarian, experimentalist, engineer, writer, and auditor.
• Integrated domain knowledge and skills — brings together brain-science knowledge, methodological skills, analysis procedures, writing conventions, and tool interfaces.
• Auditor Agent for scientific reliability — reviews claims, evidence chains, citations, hallucination risks, omitted information, and unsupported reasoning.
• Graph of Trace — records task structure, agent actions, evidence flow, and decision points as an inspectable research process.
• Extensible research tool ecosystem — connects model providers, MCP tools, literature databases, code-execution environments, and custom research utilities.
• Fast local start — install the CLI and begin working in the browser without requiring Docker.

Install

npm install -g @brainpilot/app@0.1.0
brainpilot up

Node.js 22 or later is required. Configure a supported model provider API key, or use BP_MOCK=1 for a no-key smoke run.

Published artifacts

• npm — @brainpilot/app, @brainpilot/backend-core, @brainpilot/protocol, @brainpilot/runtime, @brainpilot/skills, and @brainpilot/web at version 0.1.0.
• Docker sandbox — ghcr.io/neuroaihub/brainpilot-sandbox:0.1.0.
• GPU sandbox — ghcr.io/neuroaihub/brainpilot-sandbox-gpu:0.1.0.

See the README for architecture, case studies, setup, providers, MCP integration, and deployment options.