Doorman
3.0 AI governance is here

One gateway for your team’s AI and APIs.

Give people AI with personal keys and allowances, put your APIs behind one set of policies, and see every request and every dollar. In your own infrastructure.

Run the demoSelf-host from GitHub
Self-hosted
your cloud or VPC
Apache 2.0
no license key
AI + every API
REST, GraphQL, gRPC, SOAP

00The dashboard

One dashboard for keys, spend, and APIs.

Doorman AI usage dashboard showing spend by model over time

One gateway for

  • REST
  • GraphQL
  • gRPC
  • SOAP
  • OpenAI
  • Anthropic
  • Gemini
  • Azure OpenAI

Runs in your infrastructure

Your cloud account or private network. Provider keys, prompts, and traffic stay with you.

Open source, Apache 2.0

The whole gateway and dashboard, with no license key and no limits on traffic.

One policy model

The same users, groups, roles, and logs for AI providers and for your APIs.

01Features

Everything between your people and what they call.

Built into the gateway, not bolted on. No extra services, no per-seat pricing.

Personal AI keys

01

Everyone gets their own key. Revoke one without touching anyone else.

laptopdk_pXKZ…active
ci-runnerdk_8Hq2…active
old-macdk_Lr0w…revoked

Allowances

02

Per person, team, or tier, in dollars, tokens, or requests. Block, warn, or downgrade.

Engineering$1,840 / $2,500
Design$310 / $800
Support$500 / $500

Support · 429 AI004 · resets Dec 1

Spend by model

03

See who spent what on which model, and export it for chargeback.

claude-sonnet-5-5$1,212
gpt-4.1-mini$684
gemini-flash$97
llama-localunpriced

Roles and groups

04

One set of users, roles, and groups across AI providers and APIs.

Dana LeeAdmin ▾
Sam OrtizDeveloper ▾
Lee ParkAnalyst ▾
ci-botService ▾

Rate limits and credits

05

Throttle callers per API, sell credits, and return clean 429s.

payments/v1100 / min200
orders/v31k / min200
partner-acme20 / min429

Guardrails

06

Redact emails, card numbers, and secrets before a prompt leaves your network.

summarise this thread from

[EMAIL] using key

[AWS_KEY] and …

2 redactions · prompt not logged

AI governanceAPI management

02How it works

Change the base URL. Keep your SDK.

Callers point at Doorman with their own key. Every request is identified, checked against its limits, routed, and recorded, for AI providers and APIs alike.

python · any OpenAI SDK
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://gw.example.com/api/ai/corp-openai/v1/v1",
    api_key=os.environ["DOORMAN_AI_KEY"],  # personal dk_ key
)

# Over the team's allowance:
# 429 {"error_code": "AI004", "budget": "Engineering monthly",
#      "period_resets_at": "2026-12-01T00:00:00Z"}
Limits are documented, not hidden. AI allowances are soft limits, and we say exactly how they behave.Read the limits

03Deploy

Try it locally or run it in your cloud.

The demo runs locally with sample APIs and a built-in AI model, so you can try routing, limits, and usage without a provider key.

terminal · demo
git clone https://github.com/doorman-dev/doorman.git
cd doorman && cp .env.demo .env
docker compose -f docker-compose.yml -f docker-compose.demo.yml up --build

# in a second terminal: team allowance, employee key, 4th request blocked
./scripts/demo-ai-walkthrough.sh

Self-hosted

Free · Apache 2.0

The full gateway and dashboard in your cloud account or private network. No license key and no traffic limits.

Run the demoView on GitHub

Assisted pilot

30 days
  • Installed with you, in your environment
  • Providers and allowances for a first team
  • Written scope and capped support hours
Request a pilotHow pilots work

04FAQ

Questions teams ask first.

Can't find your answer? The documentation covers setup, operations, and the AI gateway in detail.

Read the documentation
Which APIs does Doorman handle?

REST, GraphQL, gRPC, and SOAP services, plus OpenAI, Anthropic, Gemini, Azure OpenAI, and OpenAI-compatible AI providers, all with the same users, roles, groups, and logs.

Where can Doorman run?

Locally, in a container, in your cloud account, or inside a private network. One instance runs in memory; more instances share state through MongoDB and Redis.

Do callers have to change their code?

Usually just the base URL and the key. API clients call the same paths through the gateway, and AI clients keep using the official OpenAI, Anthropic, and Gemini SDKs.

Does Doorman supply AI provider keys or resell AI?

No. You connect your own provider accounts, and each provider bills you directly. Doorman holds the keys so employees never see them.

Are AI allowances hard caps?

No, they are soft limits. Doorman checks before a request and charges after it, so requests already in flight can take spend past the limit. The AI governance page and docs describe exactly how.

Is it really free?

Yes. Doorman is Apache 2.0, including every API and AI feature. We offer paid, assisted pilots for teams that want help rolling it out.

Put one API or one team behind Doorman first.

You don't need a migration plan. Start with one service or one team's AI access, see how it runs, and grow from there. We can help you pilot it.

Run the demoRequest a pilot