ExoBrain: developer tool for Hermes Agent

Command-line reasoning engine that gives an agent a clean context window to think through hard problems

What ExoBrain does

ExoBrain adds an external reasoning step to an agent. After installing the exobrain-cli package and running exobrain init, you tell the agent in its SOUL.md or system prompt to call exobrain think for complex reasoning, planning and analysis. Each call sends a focused prompt, optionally with context from an inline string, a file or stdin, to a configured model. OpenRouter is the only built-in provider, and any OpenAI-compatible or Anthropic-compatible endpoint can be added in config.toml.

Flags set the depth (quick, normal, deep or exhaustive, each with its own token and reasoning-effort preset), a reasoning profile, JSON output, raw mode without the system prompt, token and cost statistics, and a --plan option that builds a structured plan with a session id to isolate plan state. Settings live in ~/.config/exobrain/config.toml. When you use the exobrain wrapper with Hermes, keys can come from a Hermes profile's .env file. The README also points agents to llms.txt and an optional skill. The version is 0.2.6, under the MIT license.

Key features

exobrain think command for focused reasoning calls
Depth presets from quick to exhaustive
Context from inline text, a file or stdin
Custom OpenAI-compatible and Anthropic-compatible providers
Structured plan mode with per-session plan state
Token usage and cost statistics on request

When to use it

Letting an agent offload a hard planning or analysis step to a stronger reasoning model
Piping an error log into a clean-context diagnosis
Generating structured JSON output from a reasoning call

Who it is for: Agent users who want a separate, configurable reasoning model that the main agent can call from the command line.

How it fits with Hermes Agent

The repository is tagged hermes-agent and hermes-skill, uses SOUL.md for agent instructions and can read keys from a Hermes profile .env file. It also targets OpenClaw, so it is not limited to Hermes.

How to install ExoBrain

These commands are copied from the project's README. Check the repository for the latest steps before you run them.

uv tool install exobrain-cli
pipx install exobrain-cli
pip install exobrain-cli

Requirements: An API key for at least one provider; OpenRouter is the default, or use any OpenAI-compatible or Anthropic-compatible endpoint

FAQ

What is ExoBrain?

ExoBrain is a command-line reasoning engine for agents. The agent runs exobrain think with a prompt, and ExoBrain answers using a configured model in a clean context window.

Does ExoBrain work with Hermes Agent?

It is tagged for Hermes Agent and fits its setup: you add a short instruction to the agent's SOUL.md, and with the exobrain wrapper keys can come from a Hermes profile .env file. The README also tags OpenClaw.

How do I install ExoBrain?

Run uv tool install exobrain-cli, or use pipx install exobrain-cli or pip install exobrain-cli. Then run exobrain init to set your provider key and default model.