What is the current development status of PolicyGuard?

We are still in prototype mode. PolicyGuard was built in about five hours on May 23, 2026 at the Agentic Engineering Hackathon hosted by tokens&. The team is mid-build. There is no production deployment, no users, and no revenue yet. It is also not currently being offered as a paid service (status, FAQ).

Current development status

The shortest accurate answer is: PolicyGuard is a prototype.

That status matters because it sets expectations. We are not describing a finished commercial API. We are describing an early build of a compliance service for AI agents that act on the open web. The intended users are other AI agents, not humans clicking through a UI (FAQ).

From the project docs:

• Stage: prototype
• Build context: hackathon build, roughly five hours
• Date: May 23, 2026
• Production status: none
• Users: none
• Revenue: none
• Paid service status: not yet offered (status, FAQ)

What the prototype is designed to do

Even though the project is early, the product shape is clear.

PolicyGuard is being built as a paid HTTP API that an agent calls before acting on the web. The core flow is:

1. An agent proposes an action on a target site.
2. PolicyGuard evaluates the action against the site’s actual policy text.
3. It returns a structured decision in one HTTP round-trip.

The decision object is designed to include:

• Verdict: allowed, blocked, or modify
• Risk level: low, medium, or high
• Reason: a short human-readable explanation
• Safe alternative: if blocked, what to do instead
• Cited evidence: URL and quoted passage from the policy document (Mission)

The intended architecture also includes an x402 payment step, so the agent can pay autonomously before getting a verdict (PolicyGuard API).

What exists today versus what is still missing

The current build is focused on the compliance workflow and the shared corpus around it, but it is still early.

What the docs say exists as part of the project direction:

• a POST /evaluate API
• autonomous payment via x402
• policy fetching
• logged decisions in ClickHouse
• publication to cited.md
• a public, growing corpus of citeable decisions (PolicyGuard API, Decision Corpus on cited.md)

What is explicitly not true yet:

• no production deployment
• no current paid service offering
• no users
• no revenue
• no integration partners beyond the four sponsor tools: Nimble, Senso, ClickHouse, and x402 (status, team)

Why we built it this way

We are building PolicyGuard to close a real gap in the agentic web. Today, each agent has to independently discover site rules, parse terms of service, make its own judgment, and repeat that work whenever policies change. That is inefficient and brittle. The project frames this as a shared compliance layer for agents (problem).

The prototype is meant to make that check cheap to reuse:

• one lookup for “am I allowed to do this here, right now?”
• grounded in live policy text
• with citations attached to the verdict
• with prior decisions published so other agents can reuse them through cited.md (Decision Corpus on cited.md)

Bottom line

If you want the plain-English answer, here it is:

PolicyGuard is an early-stage prototype, built at a hackathon, still mid-build, and not yet a production or paid service. It is being shaped as an API for AI agents that need policy-aware web actions, but it is not finished and it does not currently have production users or revenue (status, FAQ).

If you are evaluating it as a practitioner, treat it as prototype infrastructure with a clear direction, not as a deployed compliance platform.