Self-Hosted Docker Compose

git clone https://github.com/future-agi/future-agi.git
cd future-agi

This tutorial installs the default Standalone setup. It pulls published images, futureagi/standalone plus Postgres and ClickHouse, about 800 MB in all; there’s no source build.

You should see a future-agi/ directory with docker-compose.yml and .env.example at the root.

./bin/install

The installer checks the Docker VM’s memory and the host ports, writes .env with this install’s generated secrets, pulls the images, starts the stack and waits until it answers. Before it starts, it asks for your first account: an email, a name and a password of at least 8 characters that is not all digits and not a common password. Press Enter at the email prompt to skip and sign up in the UI later.

Just before the account prompt, the installer shows what deployment telemetry sends: once the first account exists, the install registers with api.futureagi.com, including the email addresses of owner, admin, staff and superuser accounts, then sends usage counts every 6 hours. To opt out, run ./bin/install --no-telemetry instead: only a minimal registration without email addresses is then sent.

The first boot sets up the databases and takes a few minutes; on a slow host, its migrations alone can take over half an hour. Meanwhile http://localhost:3000 shows a Future AGI is starting page with the current phase. To follow along:

docker compose logs -f app

The log ends with Future AGI is ready in .... When the installer finishes, it prints the endpoints:

You should see three containers, app, postgres and clickhouse, in docker compose ps.

See Installation for every installer flag.

Tracing needs no provider key: the example below calls OpenAI directly with the key you exported. To use the LLM gateway, the playground or LLM-based evals in the app, add a key to .env:

OPENAI_API_KEY=sk-...

Then apply it:

docker compose up -d

Email is optional too. Without it, messages are written to the app log instead of being sent. To send invites and password resets by email, add Mailgun credentials (MAILGUN_API_KEY, MAILGUN_SENDER_DOMAIN) the same way.

See Environment variables for every setting.

Open http://localhost:3000. A fresh install first shows the first-run setup screen, which checks your services. Then sign in with the account the installer created, or sign up if you skipped that prompt.

Open Keys in the sidebar (http://localhost:3000/dashboard/keys) and copy the API key and the secret key.

You should see an API key and a secret key. Update the FI_API_KEY and FI_SECRET_KEY values you exported in Install with them.

docker compose exec app python manage.py reset_password --email you@example.com

Point the instrumentation SDK at your local collector with FI_BASE_URL. Everything else is identical to the cloud setup.

import os
from fi_instrumentation import register, FITracer
from fi_instrumentation.fi_types import ProjectType
from traceai_openai import OpenAIInstrumentor
from openai import OpenAI

client = OpenAI()

# FI_BASE_URL (exported in Install) points the exporter at the local
# trace collector instead of the Future AGI cloud endpoint.
trace_provider = register(
    project_type=ProjectType.OBSERVE,
    project_name="local-stack-smoke-test",
)
OpenAIInstrumentor().instrument(tracer_provider=trace_provider)
tracer = FITracer(trace_provider.get_tracer("local-stack-smoke-test"))

@tracer.agent(name="smoke_test_agent")
def smoke_test_agent(question: str) -> str:
    response = client.chat.completions.create(
        model="gpt-4o-mini",
        messages=[{"role": "user", "content": question}],
    )
    return response.choices[0].message.content

print(smoke_test_agent("What's our policy on refunds after 30 days?"))
trace_provider.force_flush()

Open Tracing in the sidebar and select local-stack-smoke-test. You should see one parent span (smoke_test_agent) with the OpenAI call nested underneath. If the trace shows up, the collector, ClickHouse and the UI are wired correctly.