Agent intent
A machine states the bounded job it wants to perform.
Configure your agent or custom skill free. When the plan is ready, pay $29 once through Stripe; the verified webhook unlocks the exact bundle and download confirmation.
Give your agent a purpose. Choose its capabilities. Set the boundaries. Bring the build to life with CrossingKey's governance and a founder's attention to detail.
Learn → configure → review → build
Inspect a capability's purpose, data handling, and permissions before adding it to your agent's brief. Catalog previews do not execute paid tools.
Loading the capability catalog…
CrossingKey MCP is presented here as a governed abstraction layer around the x402-style payment flow: a machine can discover a capability, meet a payment challenge, and receive a receipt without confusing payment with permission to execute.
A machine states the bounded job it wants to perform.
Scope, domains, limits, approvals, and evidence are made explicit.
The capability returns a payment challenge before a paid request is fulfilled.
The customer-authorized payment is verified and paired with a receipt.
Run the facilitator preview to inspect the messages. It does not connect to a wallet, charge a card, or submit an on-chain transaction.
{
"state": "ready",
"note": "Simulation only"
}
Read the protocol foundation at x402-foundation/x402 ↗ and inspect the public CrossingKey MCP listing at Glama ↗.
Create a concrete brief with limits, permissions, and acceptance criteria. Download it as a .xkey specification or bring it to the Navigator for implementation.
Turn a plain-language brief into a governed agent, skill, workflow, application, or business tool. The builder produces an inspectable plan before anything can execute.
Configuration and review are free. When you are satisfied, sign in, pay through Stripe, and this exact build becomes downloadable as a runnable customer-owned package. Bring your own model key; external tools and spending remain disabled by default.
Digital sales are final where permitted by law. A Stripe-confirmed refund revokes entitlement and future downloads.
Save custom skill specifications in this browser, create versions, run their declared acceptance checks, and export the result. Secrets are never stored here.
Every offer begins with a defined scope, acceptance criteria, governance boundary, and delivery path. Payment never silently authorizes execution or grants access to private CrossingKey systems.
Review a proposed skill or agent for scope, permissions, sensitive-data exposure, failure modes, and test requirements.
Quote after intakeRequest review ↗Fixed-scope implementation of an approved AI workflow, governed agent, API integration, or developer tool with validation and documentation.
$1,500 USD / fixed scopePay after scope approval ↗For businesses that need a larger operating system: custom skills, approved integrations, testing, workspace records, and release controls.
Scope-basedStart a platform scope ↗Support the continued development of governed AI infrastructure, public research, and open machine-commerce work.
Optional supportSupport the work ↗Custom skill export is automatically fulfilled after Stripe confirms payment. Scoped implementation work remains a separate, manually scoped service.
Sign in to sync a brief to Supabase. A local browser buffer is always available so you can experiment with HAAR gates and step limits without a mandatory login.