An ambitious destination. A deliberate sequence.
From a gateway
to a shared language.
Agents need a consistent answer to three questions: Who allowed this? What were the limits? What actually happened?
Make every consequential action accountable.
Our long-term thesis is a portable contract for agent authorization and execution evidence, across gateways, clients, and clouds.
The baseline
Available / Developer preview
Make the boundary tangible.
An interactive policy workbench, JSON payload projection, exportable policies and simulation receipts, a public read-only MCP endpoint, and a single-process local gateway with tool allowlists, budgets, and circuit breaking.
The hosted private beta
Proposed next phase. Dates follow validation, not a feature countdown.
Checkpoint / First 10 active teams
One production workflow, end to end.
Managed login and workspaces. Secure upstream credential storage. Durable rate limits and audit metadata. One tested route from an engineering agent to an issue tracker. Self-serve connection checks and policy explanations.
Success means a new team routes its first verified call in under ten minutes, returns the following week, and can explain a blocked action without founder support.
Checkpoint / Repeat usage and paid conversion
Pay for operating confidence.
Proposed free plan: two servers and 1,000 tool calls a month. Proposed Pro plan: $79 a month, ten servers, 100,000 calls, and seven days of metadata. Default hard caps, explicit overage opt-in, and clear usage meters.
No checkout or hosted entitlement is active today. Pricing will be validated against actual support, telemetry, and data-transfer costs.
The enterprise path
Checkpoint / Funded design partners
Identity and evidence that travel together.
Per-agent and delegated user identity. Managed SSO. Exact-action approvals with durable state and expiry. Signed policy revisions, tool schema change review, and metadata export to the customer’s observability system.
SSO, private deployment, and compliance obligations need a commercially sustainable contract. An independent audit precedes any SOC 2 assurance claim.
Checkpoint / Repeatable enterprise delivery
Your data plane. One control plane.
A customer-owned execution runtime pulls signed policy bundles and sends minimal evidence outbound. Support one deployment target first, with automated upgrade checks and a documented support window.
Private-network access and emergency revocation are explicit design requirements. Arbitrary on-premise environments and bespoke forks are not an initial support promise.
A three-to-five-year thesis
The opportunity is broader than owning the route. Publish a small, vendor-neutral policy and execution receipt contract that can be evaluated and verified independently.
- Year one: earn adoption with an excellent developer workflow and dependable hosted governance.
- Year two: establish reusable policy packs, schema-drift checks, and deterministic replay of authorization decisions.
- Years three to five: work toward cross-gateway verification, customer-owned execution, partner adapters, and independent conformance tests.
These are strategic hypotheses, not promised milestones. A large footprint could come from many developers embedding an open verifier while organizations pay for fleet management and evidence delivery.
A strategic acquisition becomes plausible when an identity, observability, or API infrastructure platform can buy distribution and a trusted contract it would take years to reproduce. A valuation outcome cannot be guaranteed.
Built for leverage
- Use the official protocol SDK and managed identity. Do not maintain a private protocol or custom OAuth server.
- Keep the synchronous execution path small. Move billing, exports, and policy distribution to durable background work.
- Cache explicitly safe reads by exact inputs and authorization scope. Never use approximate matches to authorize or replay a write.
- Export evidence by default, not sensitive payloads. Customer-owned telemetry is a product feature, not a storage cost trick.
- Automate integration tests, release checks, and migrations before adding supported connectors.
The full strategy, feature mechanics, cost model, and implementation checkpoints live in the product strategy.