How do we roll this out without breaking production?

Start in verify mode: the gate observes, records verdicts, and blocks nothing while you tune the policy set against real traffic. Flip to full mode when the refusals you see are the refusals you want. Upstream failures are handled with configurable fallback routing, and the latency budgets — under 50ms enforcement, under 20ms telemetry — are regression-gated in CI.

Who owns the policies once it’s deployed?

Not you, ideally. The dashboard’s no-code editor emits the exact DSL, so GRC and security write and approve policies while platform owns the deployment. The separation is real: a policy change is a reviewed bundle rollout, not a pull request against your middleware.

What does day-2 operation look like?

Docker Compose or Helm, a distroless Go proxy with a self-probing healthcheck, and honest failure semantics: if the control plane goes down the proxy keeps enforcing the last compiled bundle and spools telemetry durably until the engine returns. It’s v0.1 — APIs may move between releases, and release notes say so.

For platform teams

One base-URL change. Governance someone else can own.

You built the LLM platform, so every compliance question now lands on you. AnchorGate turns governance into infrastructure you deploy once: policy enforcement, redaction, and evidence at the gateway — with a no-code editor so GRC owns the policies and you own the uptime.

Deploy the stack Read the architecture

$ export OPENAI_BASE_URL=http://localhost:8080/v1

The problem

Why governance keeps landing on your team

You became the compliance team

Whoever runs the gateway gets the audit questions. Without enforcement at the boundary, every ask turns into custom middleware, another sprint, and a policy implemented five slightly different ways.

Guardrails sprawl

A guardrails library per service means every team pins a different version with different rules — and coverage ends at the apps someone remembered to instrument. Infrastructure covers everything that routes through it.

The latency veto

Any governance layer that adds hundreds of milliseconds gets quietly removed by Friday. Enforcement has to live inside a budget you can put in an SLO — and the budget has to be enforced in CI, not in a blog post.

What the gate takes off your plate

Policy Enforcement

A deterministic policy DSL — IF condition THEN redact, alert, or block — compiled, human-approved, and evaluated before a request leaves. No LLM judging LLMs: the same input always gets the same verdict.

Learn more →

PII Redaction

SSNs, emails, Luhn-validated card numbers, and phone numbers stripped inline — before the request reaches any model provider. Responses and streamed output are scanned too.

Learn more →

MCP Security

Deny-by-default: exact-string target allowlists (wildcards deliberately unsupported), shell invocations rejected outright, and single-use task-bound tokens that burn on first use.

Learn more →

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 →

Design partners

No borrowed logos. A standing offer instead.

AnchorGate is v0.1. We don’t have testimonials yet, and we won’t invent them — what we have is a program for the first teams who put real traffic through the gate. We’d rather build with five serious teams than sell to fifty.

Who it’s for

Platform and GRC teams facing an EU AI Act, ISO/IEC 42001, or SOC 2 conversation this year
Agentic workloads — RAG, copilots, MCP tool use — running in regulated environments
Teams that want enforcement and evidence from the same gate their traffic already crosses

The exchange

You get

A direct line to the maintainers, not a ticket queue
Deployment support in your VPC
Your governance requirements shaping the roadmap

We get

Your ugliest governance requirements, stated plainly
Feedback from a real deployment — not a demo

Email hello@anchorgate.ai

FAQ

What platform teams ask us

How do we roll this out without breaking production?: Start in verify mode: the gate observes, records verdicts, and blocks nothing while you tune the policy set against real traffic. Flip to full mode when the refusals you see are the refusals you want. Upstream failures are handled with configurable fallback routing, and the latency budgets — under 50ms enforcement, under 20ms telemetry — are regression-gated in CI.
Who owns the policies once it’s deployed?: Not you, ideally. The dashboard’s no-code editor emits the exact DSL, so GRC and security write and approve policies while platform owns the deployment. The separation is real: a policy change is a reviewed bundle rollout, not a pull request against your middleware.
What does day-2 operation look like?: Docker Compose or Helm, a distroless Go proxy with a self-probing healthcheck, and honest failure semantics: if the control plane goes down the proxy keeps enforcing the last compiled bundle and spools telemetry durably until the engine returns. It’s v0.1 — APIs may move between releases, and release notes say so.

Deploy governance like you deploy everything else.

Clone the repo, `make up`, and put a real request through the gate this afternoon. One base-URL change per workload after that.

View on GitHub

or email hello@anchorgate.ai