Skip to main content

​
Self-Hosted Deployment

Run the complete smoltbot stack on your own infrastructure for full data residency control. All traces, integrity checkpoints, and agent data stay within your accounts. The stack runs on Cloudflare Workers (gateway, observer, API) and Supabase (Postgres database with row-level security).
For enterprise self-hosting with SLA guarantees, license keys, and dedicated support, contact us. Enterprise licenses include hybrid analysis mode, SSO/SAML integration, and priority support.

​
Prerequisites

Before starting, ensure you have:
  • Node.js 20+ and npm installed
  • A Supabase project (or any Postgres 15+ database)
  • A Cloudflare account with Workers enabled
  • A Cloudflare AI Gateway configured (for request logging)
  • An Anthropic API key (required for AIP analysis, which uses Claude Haiku)
  • Optional: OpenAI and Gemini API keys for multi-provider tracing
1

Clone the repositories

Clone smoltbot and both protocol SDKs:
git clone https://github.com/mnemom/smoltbot.git
git clone https://github.com/mnemom/aap.git
git clone https://github.com/mnemom/aip.git
2

Set up the database

Create a Supabase project (or provision your own Postgres instance) and apply the schema:
cd smoltbot/database
psql $DATABASE_URL < schema.sql
The schema creates tables for agents, traces, integrity checkpoints, drift alerts, enforcement records, and the conscience timeline. Row-level security policies ensure agents can only access their own data.
If using Supabase, you can run the schema via the SQL Editor in the Supabase dashboard. Navigate to SQL Editor, paste the contents of schema.sql, and execute.
3

Configure Cloudflare AI Gateway

The smoltbot gateway routes LLM requests through Cloudflare AI Gateway for logging and analytics. Set up an AI Gateway in your Cloudflare dashboard:
  1. Go to AI > AI Gateway in the Cloudflare dashboard
  2. Create a new gateway (e.g., smoltbot-gateway)
  3. Note the gateway ID and account ID — you will need these for worker configuration
4

Deploy the API worker

The API worker serves agent data, traces, integrity scores, and the conscience timeline.
cd smoltbot/api
npm install
Configure wrangler.toml with your Supabase credentials, then deploy:
wrangler secret put SUPABASE_URL
wrangler secret put SUPABASE_SERVICE_KEY
wrangler secret put ANTHROPIC_API_KEY
wrangler deploy
Note the deployed URL (e.g., https://api.your-domain.com).
5

Deploy the Gateway worker

The gateway worker intercepts LLM API calls, attaches tracing metadata, and proxies requests to providers.
cd smoltbot/gateway
npm install
Configure secrets and deploy:
wrangler secret put SUPABASE_URL
wrangler secret put SUPABASE_SERVICE_KEY
wrangler secret put CF_AI_GATEWAY_ID
wrangler secret put CF_ACCOUNT_ID
wrangler secret put ANTHROPIC_API_KEY
wrangler deploy
Note the deployed URL (e.g., https://gateway.your-domain.com).
6

Deploy the Observer worker

The observer is a scheduled worker that processes AI Gateway logs, builds AP-Traces, runs AAP verification, and performs AIP integrity checks.
cd smoltbot/observer
npm install
Configure secrets and deploy:
wrangler secret put SUPABASE_URL
wrangler secret put SUPABASE_SERVICE_KEY
wrangler secret put CF_AI_GATEWAY_ID
wrangler secret put CF_ACCOUNT_ID
wrangler secret put ANTHROPIC_API_KEY
wrangler deploy
The observer runs on a cron schedule (configured in wrangler.toml) to process new gateway logs.
7

Configure the CLI

Point the smoltbot CLI to your self-hosted gateway:
npm install -g smoltbot
smoltbot init --gateway=https://gateway.your-domain.com
The --gateway flag tells the CLI to route API calls through your infrastructure instead of the Mnemom-hosted gateway.
8

Verify the deployment

Check that everything is connected:
smoltbot status
Expected output
Agent: agent-xxxxxxxx
Status: active
Providers: anthropic
Gateway: https://gateway.your-domain.com (healthy)
Traces: 0
Integrity: no checkpoints yet
Make a test API call through the gateway to generate a trace:
curl https://gateway.your-domain.com/anthropic/v1/messages \
  -H "x-api-key: $ANTHROPIC_API_KEY" \
  -H "content-type: application/json" \
  -d '{
    "model": "claude-haiku-4-5-20251001",
    "max_tokens": 256,
    "messages": [{"role": "user", "content": "Hello"}]
  }'
Wait for the observer to run (check the cron schedule in observer/wrangler.toml), then verify traces were generated:
smoltbot logs

​
Architecture overview

When self-hosted, the full stack runs across your Cloudflare and Supabase accounts: 
Your App
  │
  ▼
Gateway Worker (Cloudflare)
  │ ── proxies to ──▶ Anthropic / OpenAI / Gemini
  │
  ▼
Cloudflare AI Gateway (logging)
  │
  ▼
Observer Worker (Cloudflare, scheduled)
  │ ── builds AP-Traces
  │ ── runs AAP verification
  │ ── runs AIP integrity checks
  │
  ▼
Supabase (Postgres)
  │
  ▼
API Worker (Cloudflare)
  │
  ├──▶ CLI (smoltbot)
  └──▶ Dashboard (optional, self-hosted)

​
Required secrets per worker

SecretGatewayObserverAPI
SUPABASE_URLYesYesYes
SUPABASE_SERVICE_KEYYesYesYes
CF_AI_GATEWAY_IDYesYesNo
CF_ACCOUNT_IDYesYesNo
ANTHROPIC_API_KEYYesYesNo
Never commit secrets to version control. Use wrangler secret put to securely store them in Cloudflare, or use Cloudflare’s environment variable encryption. The Supabase service key has full database access — treat it accordingly.

​
Custom domain setup

To serve the gateway and API on custom domains:
  1. Add your domain to Cloudflare DNS
  2. In the Cloudflare dashboard, go to Workers & Pages > your worker > Settings > Triggers
  3. Add a custom domain (e.g., gateway.your-domain.com)
  4. Repeat for the API worker (e.g., api.your-domain.com)

​
Next steps