A REST API for managing Hummingbot trading bots across multiple exchanges, with AI assistant integration via MCP.
Why we recommend Tailscale for production
Hummingbot API controls real trading: orders, balances, bots, and stored exchange keys. That has always required strong passwords and careful configuration—but the risk surface has grown. Tools like MCP, Condor agents, and other AI assistants make powerful API actions easier to trigger, while cloud VPSes are constantly scanned for open ports like 8000.
Tailscale is one safeguard you can add: it puts the API on a private encrypted network so only your devices can reach it, without publishing port 8000 to the internet. It does not replace proper security—use strong API and config passwords, keep exchange keys protected, and avoid exposing sensitive services publicly. Tailscale also works when the API and clients run on the same machine. But if your only client is co-located too (e.g. Condor deploying this API for itself), it already reaches you over
localhostregardless — this API's own tailnet node is only needed for other devices to reach it directly.
Recommended (Docker): install from an empty directory with the Hummingbot Deploy helper script. Docker must be installed and running first.
For VPS or remote deployments, prepare Tailscale first:
- Create a free account at tailscale.com
- Generate a reusable auth key at Settings → Keys (starts with
tskey-auth-) - Enable MagicDNS in the Tailscale admin console
Full walkthrough: hummingbot.org Tailscale guide · Securing Condor and Hummingbot API with Tailscale
curl -fsSL https://raw.githubusercontent.com/hummingbot/deploy/main/setup.sh | bash -s -- --hummingbot-apiThe script clones hummingbot-api, runs make setup (creates .env), pulls Compose images, and runs make deploy, which starts the API, PostgreSQL, and EMQX containers.
The setup script prompts for:
- Credentials — API username/password (HTTP Basic Auth) and config password (encrypts bot credentials)
- Tailscale — answer
ywhen asked Use Tailscale for secure private networking? and paste your auth key (default hostname:hummingbot-api)
If the script finishes but services did not start, run:
cd hummingbot-api
make setup
make deploy| Where you connect from | URL |
|---|---|
| Same machine as the API | http://localhost:8000 |
| Another device on your tailnet (Condor, MCP, browser) | http://hummingbot-api:8000 |
Use your API username and password for all requests. Do not open port 8000 on your public firewall when Tailscale is enabled.
| Command | Description |
|---|---|
make setup |
Create .env file with configuration |
make deploy |
Start all services (API, PostgreSQL, EMQX) |
make stop |
Stop all services |
make run |
Run API locally in dev mode |
make install |
Install conda environment for development |
make build |
Build Docker image |
make tailscale-status |
Show Tailscale connection + serve (proxy) status |
make doctor |
Verify dependencies, .env, containers, port exposure and API access |
After hummingbot-api is running, these services are available:
| Service | Local URL | Tailnet URL (when Tailscale enabled) | Description |
|---|---|---|---|
| API | http://localhost:8000 | http://hummingbot-api:8000 | REST API |
| Swagger UI | http://localhost:8000/docs | http://hummingbot-api:8000/docs | Interactive API documentation |
| PostgreSQL | localhost:5432 | — | Database |
| EMQX | 127.0.0.1:1883 | — | MQTT broker (auth required, loopback-only) |
| EMQX Dashboard | http://127.0.0.1:18083 | — | Broker admin (admin / BROKER_DASHBOARD_PASSWORD) |
Broker access. The broker requires a username and password and is published on the loopback interface only — bot containers run with
network_mode: hostand reach it at127.0.0.1:1883, while the API reaches it in-network asemqx:1883. Credentials come fromBROKER_USERNAME/BROKER_PASSWORDin.env;make deployseeds them into the broker viamake emqx-auth. To change them afterwards, edit.envand runmake emqx-auth-reset— EMQX only imports the bootstrap file for accounts it does not already have.The dashboard login is a separate credential,
BROKER_DASHBOARD_PASSWORD— deliberately not the same value asBROKER_PASSWORD, since that one is written into every bot instance'sconf_client.ymland the dashboard grants full broker admin (rules, connectors), not just the scoped MQTT accessemqx/acl.confgivesBROKER_PASSWORD. Like the MQTT bootstrap account, EMQX only applies it on first init, so changing it in.envalso needsmake emqx-auth-reset(it wipes the sameemqx-datavolume, rotating both passwords together).The broker also denies every topic outside
hbot/#andhummingbot-api/response/#(emqx/acl.conf), so a leaked broker credential cannot be used to read the whole bus or to drive the rule engine. Runmake emqx-auditto print the broker's listeners, auth, authorization and any rules, actions, connectors or bridges — a rule nobody added can make the broker issue authenticated HTTP requests into internal services, and it survives restarts. Usemake emqx-audit EMQX_CONTAINER=<name>to check another deployment.
Ports. The API (
8000) and Postgres (5432) also bind to127.0.0.1by default. SetAPI_BINDin.envif something off-box must reach the API — prefer a specific interface over0.0.0.0; with the Tailscale overlay,API_BIND=<tailscale-ip>keeps MagicDNS working without publishing the API to the internet.
PostgreSQL and EMQX are bound to 127.0.0.1 only — they're never reachable
from another machine, on the tailnet or otherwise. Bot containers still reach
them fine (they run with network_mode: host, so 127.0.0.1 is their host's
loopback too).
Production: use
http://hummingbot-api:8000(MagicDNS) instead oflocalhostwhen MCP runs on a different device than the API. Both must be on the same Tailscale account.
claude mcp add --transport stdio hummingbot -- \
docker run --rm -i \
-e HUMMINGBOT_API_URL=http://hummingbot-api:8000 \
-v hummingbot_mcp:/root/.hummingbot_mcp \
hummingbot/hummingbot-mcp:latestFor local-only dev on the same machine, use http://host.docker.internal:8000 instead.
Then use natural language:
- "Show my portfolio balances"
- "Set up my Binance account"
- "Create a market making strategy for ETH-USDT"
Add to your config file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json
{
"mcpServers": {
"hummingbot": {
"command": "docker",
"args": ["run", "--rm", "-i", "-e", "HUMMINGBOT_API_URL=http://hummingbot-api:8000", "-v", "hummingbot_mcp:/root/.hummingbot_mcp", "hummingbot/hummingbot-mcp:latest"]
}
}
}Restart Claude Desktop after adding.
Gateway enables decentralized exchange trading. Start it via MCP:
"Start Gateway"
Or via API at http://localhost:8000/docs using the Gateway endpoints (POST /gateway/start with an empty body).
Gateway always runs secured: it serves TLS + mutual-cert (mTLS) authentication and the required
certificates are auto-generated on first start under bots/gateway-files/certs/ — no passphrase
prompt and no manual cert step. The single secret protecting the Gateway (TLS + wallet encryption)
is your CONFIG_PASSWORD. Once running, Gateway is available at https://localhost:15888.
There is no development/insecure mode: a Gateway holding wallet keys must never be served over plain HTTP, so the API only ever runs it with mTLS.
The .env file contains all configuration. Key settings:
USERNAME=admin # API username
PASSWORD=admin # API password
CONFIG_PASSWORD=admin # Encrypts bot credentials
DATABASE_URL=... # PostgreSQL connection
GATEWAY_URL=... # Gateway URL (for DEX)
# Tailscale (recommended for production)
TAILSCALE_ENABLED=true
TAILSCALE_AUTH_KEY=tskey-auth-...
TAILSCALE_HOSTNAME=hummingbot-api # MagicDNS hostname on your tailnet
# Published-port bind addresses (both default to 127.0.0.1 in docker-compose.yml).
# Widen API_BIND only if something off-box must reach the API directly — with the
# Tailscale overlay, API_BIND=<tailscale-ip> keeps MagicDNS working without ever
# publishing the API to the internet. See docker-compose.yml / docker-compose.tailscale.yml
# API_BIND=127.0.0.1
# DB_BIND=127.0.0.1Edit .env and restart with make deploy to apply changes.
Tailscale creates a private WireGuard network (tailnet) that makes the API accessible only to devices on your tailnet — no open ports, no firewall rules needed.
Use this when running on a VPS or cloud server and want to access the API privately from another machine (e.g. Condor or MCP tools).
- Create a free account at tailscale.com
- Go to Settings → Keys: tailscale.com/admin/settings/keys
- Click Generate auth key — check Reusable for multiple deployments
- Copy the key (starts with
tskey-auth-) - Enable MagicDNS in the Tailscale admin console
Run make setup and answer y when prompted:
Use Tailscale for secure private networking? [y/N]
This adds the following to .env:
TAILSCALE_ENABLED=true
TAILSCALE_AUTH_KEY=tskey-auth-...
TAILSCALE_HOSTNAME=hummingbot-api # MagicDNS hostname on your tailnetmake deployWhen TAILSCALE_ENABLED=true, this automatically runs:
docker compose -f docker-compose.yml -f docker-compose.tailscale.yml up -dA Tailscale sidecar container joins your tailnet using network_mode: host. hummingbot-api's port 8000 is bound to loopback only by default (API_BIND in .env — see Configuration), so it does not sit on any public interface. The sidecar's tailscale serve config (tailscale-serve.json, applied automatically via TS_SERVE_CONFIG when the sidecar starts — no tailscale serve command needed) is what makes it reachable again: it forwards the tailnet IP's :8000 to 127.0.0.1:8000, so the API ends up reachable at http://hummingbot-api:8000 from any device on the same tailnet, and only from the tailnet.
Once on the same tailnet, use the MagicDNS hostname instead of localhost:
claude mcp add --transport stdio hummingbot -- \
docker run --rm -i \
-e HUMMINGBOT_API_URL=http://hummingbot-api:8000 \
-v hummingbot_mcp:/root/.hummingbot_mcp \
hummingbot/hummingbot-mcp:latestWhen TAILSCALE_ENABLED=true, make run will automatically install Tailscale if needed, connect to your tailnet, and bind uvicorn to 127.0.0.1 only (Tailscale handles external access).
make tailscale-statusThis shows both tailnet peer status and tailscale serve status — confirm
port 8000 shows up as forwarded, not just that the node joined the tailnet.
From another device on your tailnet:
curl -u YOUR_USERNAME:YOUR_PASSWORD http://hummingbot-api:8000/- Portfolio: Balances, positions, P&L across all exchanges
- Trading: Place orders, manage positions, track history
- Bots: Deploy, monitor, and control trading bots
- Market Data: Prices, orderbooks, candles, funding rates
- Strategies: Create and manage trading strategies
Full API documentation at http://localhost:8000/docs
make install # Create conda environment
conda activate hummingbot-api
make run # Run with hot-reloadStart here:
make doctorChecks dependencies, .env (including credentials still left at well-known
defaults), the hummingbot-api / hummingbot-broker / hummingbot-postgres
containers, which ports are on a public interface, Tailscale's tailnet and
serve status, whether the API actually answers an authenticated request, and
whether it is connected to the MQTT broker. Read-only, and it names the fix for
whatever it finds.
Bots deploy but report nothing — no controllers, no logs, no performance?
The API is up and REST works, but it is not connected to the broker: bot status arrives over MQTT, so a bot with no broker is a bot with nothing to say. Confirm it, then re-seed:
make doctor # "Broker connection" says whether the API is connected
docker compose logs emqx | grep -i "auth-bootstrap"
make emqx-auth-reset # rewrites the bootstrap file and re-seeds the brokerA Permission denied on /opt/emqx/etc/auth-bootstrap.csv means the broker
could not read the file it seeds the account from, so no account was ever
created and the API's correct credentials came back Not authorized. The broker
still comes up healthy either way, which is why this shows up as missing bots
rather than as a broker error.
API won't start?
docker compose logs hummingbot-apiDatabase issues?
docker compose down -v # Reset all data
make deploy # Fresh startCheck service status:
docker ps | grep hummingbotTailscale not connecting?
make tailscale-status # Check tailnet peersConfirm the node appears in tailscale status and that MagicDNS is enabled in your Tailscale admin console.
- Docs: https://hummingbot.org/hummingbot-api/
- Tailscale guide: https://hummingbot.org/hummingbot-api/tailscale/
- API Docs: http://localhost:8000/docs
- Issues: https://github.com/hummingbot/hummingbot-api/issues