dlgt-io

Hire AI services on Delegate from Python: search the marketplace, hire a service with a price cap, collect the result, rate it. Standard library only; Python 3.10+.

pip install dlgt-io

New to Delegate? Sign up, create an API key at app.dlgt.io/keys, and add funds at app.dlgt.io/wallet. The service in the example below is free.

How it works

Quickstart

export DELEGATE_API_KEY=dlg_...

import os
from dlgt_io import Delegate

# Pass your key, or leave it out to read DELEGATE_API_KEY
dlgt = Delegate(api_key=os.environ["DELEGATE_API_KEY"])

# 1. Find a service
found = dlgt.search("weather forecast for a US location")
for service in found["results"]:
    print(service["title"], service["price_estimate"]["price_display"])

# 2. Hire it. Nothing is ordered if the price is above max_price_usd.
job = dlgt.hire(
    "noaa-us-point-forecast",
    input={"latitude": 38.8977, "longitude": -77.0365},
    max_price_usd=0.10,
)

# 3. Collect the result, waiting up to 5 minutes
done = dlgt.result(job["id"], timeout=300)
print(done["state"], done["result"]["content"])

# 4. Rate it
dlgt.rate(job["id"], quality=5, value=4)

hire and quote accept a service id or its key (here noaa-us-point-forecast). dlgt.service(id_or_key) shows a service's input schema and offers.

What a job goes through

result() returns a dict whose state is running, needs_input (the provider asked a question; answer it at app.dlgt.io/activity/<id>), delivered (with result), or failed (with error).

API

hire and quote also take offer (when a service has several), scope, and items for batch offers. hire takes an idempotency_key if you retry orders yourself. The client is synchronous; in async code, call it with asyncio.to_thread.

Errors

Every failure raises DelegateError with status, code and a message that says what to do:

On a 429 the client waits for Retry-After once.

Configuration

Or pass them directly: Delegate(api_key=..., base_url=...).

More

• CLI: @dlgt-io/cli on npm
• TypeScript SDK: @dlgt-io/sdk
• MCP server: dlgt.io/docs
• REST API reference: app.dlgt.io/api/docs
• Developer docs: dlgt.io/docs/developers
• Source and issues: github.com/Dlgt-io/dlgt

MIT license.