Anda Brain Quick Start

This guide provides a minimal flow from zero to a usable state, including:

• Installing anda-cli (Priority)
• Configuring environment variables
• Starting the service (Local Build or Docker, Local File Storage)
• Generating a CWT
• Creating a Space
• Submitting Formation
• Executing Recall
• Reading conversation logs

0. Prerequisites

• macOS / Linux
• Installed tools: git, jq, curl
• If building from source: rust, cargo, go
• If using Docker: docker
• A valid LLM API Key (e.g., Gemini, MiniMax, Xiaomi Mimo)

1. Working Directory

mkdir anda-brain
cd anda-brain
mkdir db

2. Install anda-cli (Complete this first)

Choose one: Download from Releases, or build from source.

Option A: Download the executable from Releases (Recommended)

Repository Releases:

• https://github.com/ldclabs/anda-brain/releases

After downloading the anda-cli executable that matches your system:

wget -O anda-cli https://github.com/ldclabs/anda-brain/releases/download/v0.6.0/anda-cli-macos-arm64
chmod +x anda-cli

Option B: Build anda-cli locally

cd path/to/anda-brain/anda-cli
go build -o anda-cli .
chmod +x anda-cli
mv anda-cli ../anda-brain/
cd ../anda-brain/

Verify the installation:

./anda-cli --help

3. Prepare Variables and Keys

3.1 Generate Ed25519 Key (for CWT signing)

./anda-cli keygen --json > keys.json
cat keys.json

3.2 Configure General Environment Variables

Create a .env file required to run Anda Brain. Example content:

LOG_LEVEL='info'
LISTEN_ADDR='0.0.0.0:8042'
SHARDING_IDX='0'
# The public key generated in the previous step. Separate multiple public keys with commas.
ED25519_PUBKEYS='YOUR_ED25519_PUBKEYS'
# You can replace this with your own valid Principal text (e.g., your existing principal id).
MANAGERS="aaaaa-aa"
# Can be replaced with the model family you are using, e.g., 'gemini', 'openai', 'deepseek', etc.
MODEL_FAMILY='anthropic'
MODEL_NAME='MiniMax-M2.7-highspeed'
MODEL_API_BASE='https://api.minimaxi.com/anthropic/v1'
MODEL_API_KEY='YOUR_MODEL_API_KEY'

4. Start Anda Brain (Local File Storage)

Choose one: Run via local build, or run via Docker (supports pulling remote images).

Option A: Run via local build

cd path/to/anda-brain
cargo build -p anda_brain --release
mv target/release/anda_brain ../anda-brain/
cd ../anda-brain/

./anda_brain local --db ./db

If you prefer not to compile locally, you can also download the corresponding anda_brain executable for your system from the Releases page:

• https://github.com/ldclabs/anda-brain/releases

After downloading, you can start it similarly using local --db ./db.

Option B: Run via Docker (Local file storage)

B1. Pull the remote image directly (Recommended)

On Apple Silicon (M1/M2/M3) macOS, it is recommended to explicitly specify the platform to avoid the requested image's platform (linux/amd64) does not match ... warning:

export DOCKER_PLATFORM=linux/amd64

docker pull --platform $DOCKER_PLATFORM ghcr.io/ldclabs/anda_brain_amd64:latest

docker run --rm --platform $DOCKER_PLATFORM -p 8042:8042 \
	-v "$(pwd)/db:/app/db" \
  -v "$(pwd)/.env:/app/.env" \
	ghcr.io/ldclabs/anda_brain_amd64:latest local --db /app/db

B2. Build the Docker image locally

# Building an arm64 local image is recommended for Apple Silicon
docker buildx build --platform linux/arm64 -f anda_brain/Dockerfile -t anda_brain:local --load .

4.2 Verify service availability

Open another terminal and execute:

# Set environment variables first
export ANDA_CWT_KEY="$(jq -r '.private_key' keys.json)"
export ANDA_BASE_URL='http://127.0.0.1:8042'

curl -s "$ANDA_BASE_URL/info" | jq .

5. Generate CWT

5.1 Generate an Admin Token (for creating a Space)

export ANDA_TOKEN="$(./anda-cli cwt \
	--subject "aaaaa-aa" \
	--audience '*' \
	--scope '*' \
	--expiration 7200 \
	--json | jq -r '.token')"

6. Create a Space

./anda-cli admin create-space \
	--user "aaaaa-aa" \
	--space-id "demo" \
	--tier 4

If you need to create a CWT token for the space:

./anda-cli cwt \
	--subject "aaaaa-aa" \
	--audience "demo" \
	--scope '*' \
	--expiration 7200 \
	--json | jq

Create a Space token for Openclaw or other agent integrations:

./anda-cli --space-id demo management add-token --scope "*" --name openclaw

View active Space tokens:

./anda-cli --space-id demo management list-tokens

Revoke a Space token:

./anda-cli --space-id demo management revoke-token STxxx

View Space info:

./anda-cli --space-id demo info

7. Submit Formation

# The ANDA_TOKEN environment variable is used here
./anda-cli --space-id demo \
	formation --messages '[
		{"role":"user","content":"I prefer dark mode, and my timezone is UTC+8."},
		{"role":"assistant","content":"Got it, I have noted your preference and timezone."}
	]'

Formation is processed asynchronously. You can monitor the progress using formation-status or by checking the conversation list.

./anda-cli --space-id demo formation-status

View complete Formation processing logs:

anda-cli --space-id demo conversations get 1

View more Formation-related commands:

anda-cli formation --help

8. Execute Recall

Recall is processed synchronously and will wait for the final result to return:

./anda-cli --space-id demo \
	recall "What are the preferences of this user?"

View complete Recall processing logs:

anda-cli --space-id demo conversations --collection recall get 1

9. View Conversation Logs

9.1 List Conversations

List the latest 10 Formation conversation records:

./anda-cli --space-id demo \
	conversations list --limit 10

List the latest 10 Recall conversation records:

./anda-cli --space-id demo \
	conversations --collection recall list --limit 10

List the latest 10 Maintenance conversation records:

./anda-cli --space-id demo \
	conversations --collection maintenance list --limit 10

9.2 Get Details of a Single Conversation

Replace <conversation_id> with the conversation ID returned in the previous step:

./anda-cli --space-id demo \
	conversations get <conversation_id>

10. Integration

10.1 Direct use of the HTTP API

1. Memorize (Formation): Send conversations to encode memories

curl -sX POST http://localhost:8042/v1/demo/formation \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "messages": [
      {"role": "user", "content": "I work at Acme Corp as a senior engineer."},
      {"role": "assistant", "content": "Nice to meet you! Noted that you are a senior engineer at Acme Corp."}
    ],
    "context": {"counterparty": "user_123", "agent": "onboarding_bot"},
    "timestamp": "2026-03-25T10:30:00Z"
  }'

2. Recall: Query memories before generating a response

curl -sX POST http://localhost:8042/v1/demo/recall \
  -H "Authorization: Bearer $TOKEN" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Where does this user work and what is their role?",
    "context": {"counterparty": "user_123"}
  }'

10.2 Integration with OpenClaw

Installation Prompt:

Install the plugin from https://brain.anda.ai/SKILL.md:
spaceId: your_space_id
spaceToken: your_space_token
baseUrl: "http://localhost:8042"