Give any entity an @identity that AI agents can discover, understand, and use.
delphy.network/@green-monster/skill.md↗Agents read your skill.md to learn what's possible. They ask for approval before any action that touches the world.
Write a delphy.json once. Claim your @identity. Get discovered by any AI agent that can browse the web.
Define your capabilities in a single JSON file.
Claim your @identity with OAuth verification.
Any agent that can browse can now invoke you.
Businesses, people, services, APIs, workflows. Anything an agent should find gets an @identity.
Local salsa maker. Catering pickup. Katy, TX.
Full-stack builder. AI engineering for hire.
Residential real estate. Katy & west Houston.
City service requests. Pothole, trash, water.
Flights, hotels, packages. Travel booking.
Limestone supply & cuts. Monterrey, MX.
One file declares everything an agent needs to know. Self-describing, versioned, auditable.
Every capability declares its risk. Higher risk triggers human approval. Safe by design, not by hope.
Agents can read data but cannot trigger any action that affects the world.
Action that reaches a third party but does not change state on either side.
Action that mutates the entity's internal state or records.
Action that moves money, places orders, or creates a financial commitment.
Action with legal weight, contractual implications, or regulatory exposure.
Action that cannot be undone once executed. The point of no return.
Browser, search, REST, terminal, MCP, GPT. Same @identity, every channel.
Every @identity resolves to a public profile at a stable URL.
Filter the registry by kind, location, or capability.
REST endpoints for programmatic discovery and resolution.
Claim, manage, and publish from your shell.
Drop-in server for Claude, Cursor, and MCP-compatible agents.
Native discovery and resolution inside ChatGPT.
Pick a handle. Get your claim secret. Be found by every agent in minutes.