brainclaw + Hermes Agent

Autonomous skills-first agent with native MCP access.

Hermes Agent is an autonomous task-based agent from Nous Research with native MCP client support and a skills system. brainclaw configures Hermes as a Tier B autonomous surface today: machine-level MCP through ~/.hermes/config.yaml plus the shared .agents/skills/brainclaw/SKILL.md skill, so Hermes can read plans, claims, decisions, and traps from the same coordination layer as the rest of the team.

Install via Hermes Agent

Paste this prompt directly into Hermes Agent and let it handle the setup.

Please install brainclaw and initialize it for this project.

Run: npm install -g brainclaw && brainclaw init

brainclaw will detect Hermes, write AGENTS.md with the project's plans, decisions, and traps, install the shared brainclaw skill at .agents/skills/brainclaw/SKILL.md, and configure the Brainclaw MCP server in ~/.hermes/config.yaml.

Auto-setup

Run one command. brainclaw detects Hermes Agent and writes its native files automatically, including AGENTS.md.

npm install -g brainclaw && brainclaw init

Detection: HERMES_* env vars or ~/.hermes.

What gets generated

After running brainclaw init, these files are created or updated:

AGENTS.md
.agents/skills/brainclaw/SKILL.md
~/.hermes/config.yaml
.brainclaw/

MCP integration

brainclaw configures MCP in ~/.hermes/config.yaml, giving Hermes Agent dynamic access to workspace context via MCP tools.

 bclaw_work - open a session and claim scope in one call
 bclaw_context - ranked prompt-ready context
 bclaw_create / bclaw_update - capture decisions and traps, evolve plans
 bclaw_release_claim - close the claim and write a handoff 