Agentra LabsAgentra Labs DocsPublic Documentation
WebsiteGitHub

Get Started

Quickstart

Get started with AgenticCognition in under five minutes. This guide walks through installation, model creation, belief management, and verification.

Get started with AgenticCognition in under five minutes. This guide walks through installation, model creation, belief management, and verification.

Install

curl -fsSL https://agentralabs.tech/install/cognition | bash

The installer places acog and acog-mcp in ~/.agentra/bin/ and configures detected MCP clients automatically.

Verify installation

acog --version
acog-mcp --version

Both commands should print version information. If "command not found" appears, add ~/.agentra/bin to your PATH:

export PATH="$HOME/.agentra/bin:$PATH"

Create your first living user model

MODEL_ID=$(acog model create --format json | jq -r '.id')
echo "Created model: $MODEL_ID"

Add beliefs

acog belief add $MODEL_ID "I value honesty above all" --domain values --confidence 0.9
acog belief add $MODEL_ID "I prefer deep work over meetings" --domain work --confidence 0.85
acog belief add $MODEL_ID "I learn best by building things" --domain growth --confidence 0.8

Record a heartbeat

acog model heartbeat $MODEL_ID --context "initial setup and belief capture"

Generate a soul reflection

acog model soul $MODEL_ID

This produces a multi-dimensional reflection of the user based on all captured beliefs.

Check model vitals

acog model vitals $MODEL_ID --format table

Verify with MCP

Once installed, the acog-mcp binary is available as a Model Context Protocol server. Any MCP-compatible client (Claude Desktop, Cursor, Windsurf, VS Code) can use cognition tools immediately.

Test the MCP server manually:

echo '{"jsonrpc":"2.0","method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"0.1"}},"id":1}' | acog-mcp

Next steps