Govern your agents without renting governance from your model's vendor.
AgentGuard® runs as a library inside your runtime — no proxy, no data plane. Enforce spend caps before a call is dispatched, prove every agent action with a signed tamper-evident ledger, and see the spend a provider console can't: the open-weight and self-hosted dark tokens.
Prompts, provider keys, and the signing key never leave your process. Governance is a library review, not a data-vendor security review.
Your token bill doubles every 45 days. Then the CFO asks for the ROI.
Cheap tokens didn't reduce the problem — they multiplied it. Teams run parallel agents across a mix of frontier APIs, open-weight models, and self-hosted GPUs. Spend fragments; attribution disappears; nobody can prove what each agent did or spent.
You can't rent governance from a vendor you might compete with.
Frontier labs watch where value accrues on top of their models, then move in — Figma → Claude Design, Cursor → Claude Code. Your spend governance and audit trail shouldn't be owned by the vendor that may one day compete with you. And a proxy sees only its own traffic — it can't gate a cross-agent chain, and it can't see a token it never routed.
A neutral control plane, one npm install.
Pre-dispatch spend caps
Block or downgrade a call before it costs money, scoped by model, provider, agent, or workflow. The saved-by-policy number is the ROI line the CFO asked for.
See the quickstart →Signed, tamper-evident ledger
Every decision is Ed25519-signed and hash-chained in your runtime. Anyone with the public key can verify it. Nothing to certify because nothing leaves the process.
Verify a receipt →Multi-agent attestation
Gate a payment on an unbroken, signed cross-agent chain. Tamper one upstream receipt and the escalation is denied. A proxy can't do this — it sees single calls.
Watch the demo →Dark-token cost attribution
See spend by hour, model, and workload — including the self-hosted and open-weight usage no provider bill shows. Frontier-vs-dark split, computed from the signed ledger.
Read the docs →Governance that follows the model — even your own weights.
- Hot-swap frontier → open-weight → self-hosted to cut cost. The audit stays one continuous, verifiable chain — keyed to the workload, not the model.
- Works when inference is local: a proxy can't sit in the path of your own GPUs; a library can.
- Framework middleware for the AI SDK, LangChain, OpenRouter, and Claude Code — insert as a governance layer and inherit your existing stack.
- Governance for the 99% who can't build a Coinbase- or DoorDash-grade routing layer in-house.
AgentGuard® Reg. No. 8281464 · DAG Trust Attestation Tokens, Patent D §7.3 (App. No. 63/984,626), and additional U.S. provisional applications · Dunecrest Ventures Inc.