gbrain

Search and write a shared company knowledge brain via the gbrain CLI.

Updated Aug 17, 2026
One-click install
npx skills add https://github.com/AMC-JTC/gbrain-1 --skill gbrain-amc-jtc
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gbrain
Source: https://github.com/AMC-JTC/gbrain-1/tree/main/docs/integrations/qm-harness-snippets
Command: npx skills add https://github.com/AMC-JTC/gbrain-1 --skill gbrain-amc-jtc

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve? Agents and team members lose context that lives outside their current conversation. This Skill connects your sandbox to the org's central gbrain knowledge base so you can search org-wide knowledge and persist durable facts that outlive a single session. ## Core Features & Use Cases - Hybrid search and retrieval: Run gbrain search, gbrain get, and gbrain query --json to answer questions about people, projects, decisions, and history with server-side read isolation. - Write-fenced durable memory: Persist conclusions with gbrain put under your own emp-<slug>/ and chan-<channel>/ slug prefixes, and cross-reference pages with gbrain link. - Roster-driven provisioning: The included provision-scopes.sh script registers one OAuth client per employee, converges write fences to the roster, and flags offboarded credentials for revocation. - Use Case: Before answering "who decided our database choice and why", run gbrain search to pull the decision page, then write new findings back to chan-eng/decisions/ so the whole team can find them. ## Quick Start Ask the agent to search the company brain for who decided a specific technical choice and why before answering the question.

Frequently Asked Questions about gbrain

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I search the company knowledge base with gbrain?

Run `gbrain search "your question"` for hybrid semantic plus keyword search, or `gbrain query "question" --json` for agent-tuned output. Use `gbrain get <slug>` to read a single page directly.

How do I write durable knowledge to the gbrain brain?

Use `gbrain put <slug> --content "..."` under your own write-fenced prefixes such as `emp-<your-slug>/` or `chan-<channel>/`. Writes outside your bound slug prefixes are rejected server-side.

How do I set up gbrain credentials in a new sandbox?

Run `gbrain init --mcp-only` with the issuer URL, MCP URL, client ID, and client secret from the secret handoff, passing the secret via `--oauth-client-secret` so it persists to config. Verify with `gbrain whoami` before other commands.

Why does gbrain fail with 'No client_secret available'?

This happens when the secret was passed via the GBRAIN_REMOTE_CLIENT_SECRET environment variable, which is deliberately not written to the config file. Re-run init with `--oauth-client-secret` so the credential persists on disk.

Can I run gbrain remote doctor to check connectivity?

No. `gbrain remote doctor` requires admin scope, which per-employee clients do not have by design. Use `gbrain whoami` as the read-scope health check instead.

How do I provision gbrain clients for a whole team roster?

Use the included provision-scopes.sh script with a roster file listing channels and employees. It registers one OAuth client per employee with write fences, rescopes existing clients on re-runs, and flags offboarded credentials for revocation.