Developers
Schema-first, self-hosted, yours to read.
Everything here points at things that exist today: a quickstart that boots the full stack with seeded demo data, an OpenAI-compatible proxy you integrate with one base-URL change, and a typed schema you can build against. No API keys to request, no sandbox tenant, no waitlist.
# quickstart — Docker Compose with healthchecks
git clone https://github.com/anchorgate-ai/anchorgate.git
cd anchorgate
cp .env.example .env
make up
make seed
# then route a workload through the gate
export OPENAI_BASE_URL=http://localhost:8080/v1
# Kubernetes instead?
helm install anchorgate deploy/helm/anchorgate
Quickstart
A real request through the gate this afternoon
make up
brings up the proxy (:8080), the engine (:8000), and the dashboard (:3000) with
healthchecks;
make seed
loads a full governance scenario — AI systems, policies, a risk register, and demo
users — so the dashboard is interesting from minute one.
Then the integration is the base-URL change below. Everything else — enforcement, redaction, signing — happens at the gate, not in your code.
v0.1 — early and moving fast. APIs may change; we say so rather than pretend otherwise.
OpenAI-compatible
Integrate with one base-URL change
AnchorProxy speaks the OpenAI API. Keep your SDK — change the base URL, and every request now crosses the policy gate.
from openai import OpenAI
client = OpenAI(base_url="http://localhost:8080/v1") # was api.openai.com
response = client.chat.completions.create(
model="gpt-4.1",
messages=[{"role": "user", "content": "Hello through the gate"}],
)
Fronts OpenAI, Anthropic, and any OpenAI-compatible server — Ollama, vLLM — with configurable fallback between upstreams. Full quickstart →
Resources
The pointers that matter
Six links, all real. When something doesn't exist yet — a separate docs site, a cookbook section — we'd rather leave the card out than link a stub.
GitHub repository
The full stack — Go proxy, Python engine, Next.js dashboard — developed in the open. Issues and PRs welcome; it’s v0.1 and we say so.
github.com/anchorgate-ai/anchorgate ↗
Decision Event Schema
The single JSON Schema every decision record conforms to. The Go, Pydantic, and TypeScript types are generated from it and drift-checked in CI — build your tooling against the same records the dashboard reads.
des.schema.json ↗
Control-plane API
AnchorEngine is FastAPI with interactive OpenAPI docs at :8000/docs once the stack is up. Policies, assets, risks, posture, Trust Packet exports — the dashboard is a client of the same API you get.
Bring the stack up →
Helm chart
Kubernetes deploys live at deploy/helm/anchorgate in the repository, alongside the Docker Compose quickstart with healthchecks and seeded demo data.
Deploy on Kubernetes ↗
Changelog
GitHub Releases is the changelog — one honest place instead of a page that drifts. Release notes flag breaking API changes explicitly while we’re pre-1.0.
GitHub Releases ↗
Security contact
Found a vulnerability? Email security@anchorgate.ai — we respond within two business days. The disclosure policy lives on the security page.
Responsible disclosure →
FAQ
Developer questions
- Which providers and runtimes does it work with?
- AnchorProxy exposes an OpenAI-compatible surface, so integration is a base-URL change — no SDK swap. It fronts OpenAI and Anthropic upstreams and any OpenAI-compatible server, including Ollama and vLLM for local models, with configurable fallback routing between them.
- Is there an API beyond the proxy?
- Yes — AnchorEngine is a FastAPI control plane with interactive OpenAPI docs at :8000/docs once the stack is up. Policies, assets, risks, posture, and Trust Packet exports are all reachable over it; the dashboard is a client of the same API you get.
- How do I follow releases and breaking changes?
- GitHub Releases on the main repository — that’s the changelog, and we’d rather keep it in one honest place than run a separate page that drifts. It’s v0.1: APIs may change between releases, and release notes say so when they do.
Clone it, audit it, gate something.
The fastest way to evaluate AnchorGate is to put a real request through it. The stack is free, the code is open, and the seed data gives you something worth looking at in the dashboard.
View on GitHubor email hello@anchorgate.ai