Quick Start¶
Get Cognibrain running and store your first memory in under 5 minutes.
1. Install and Initialize¶
You should see output confirming the solo-dev profile is active with local storage.
2. Check Status¶
This shows runtime state, memory health, active connections, and suggested next actions.
3. Add Your First Memory¶
Memories are durable facts that agents can recall. They persist across sessions and restarts.
4. Retrieve Context for a Task¶
Cognibrain returns a compact context pack with relevant memories for the given task — your release memory will appear here.
5. Guard an Action¶
The action guard checks if any memories warn against or constrain the planned action. JSON output makes it easy for agents to parse programmatically.
6. Record Patch Evidence¶
After completing work, record what files changed, what commands ran, and what memories were relevant. This builds the audit trail.
7. Run Proof¶
The proof command summarizes the current evidence state — what's been recorded, what's verified, and what needs attention.
Full Lifecycle Example¶
Here's a complete agent lifecycle in one flow:
# Before starting work
npx cognibrain context --task "fix auth token refresh" --json
# Before a risky operation
npx cognibrain guard --action "npm test" --json
# After the operation completes
npx cognibrain outcome --command "npm test" --exit-code 0 --json
# After completing the task
npx cognibrain patch-evidence --task "fix auth token refresh" --json
# End the session
npx cognibrain session-end --json
MCP-native agents
If your agent supports MCP (Model Context Protocol), you can use the MCP integration instead of CLI commands. See MCP Integration for setup.
What's Next?¶
-
CLI Usage
Deep dive into all CLI commands and workflows.
-
Connectors
Connect GitHub, Jira, Slack, and 16+ other systems.
-
Memory Management
Learn how memories are stored, scoped, and maintained.
-
Configuration
Environment variables, profiles, and runtime settings.