Gateways
Publish your tools once, reach them from any agent or client. A gateway takes the tools you have already built and exposes them over MCP, A2A, UTCP, and Agent Skills, each from its own authenticated endpoint. You pick which tools a gateway serves, so every endpoint has a focused, purpose-built surface.

The payoff
Scope a gateway to the tools you want, generate a key on its Authentication tab, then add it to Claude Code in one line:
claude mcp add petstore --transport http https://api.almyty.com/acme/petstore
# clients send: x-api-key: <your gateway key>> look up pet #2 and show its status
Claude calls getPetById via almyty -> "Rosie", availableThe same gateway also answers over A2A and UTCP, so other agents and frameworks reach the identical tools without a rebuild.
Create one
- Open Gateways in the sidebar and click Create Gateway
- Pick a protocol (MCP, A2A, UTCP, or Skills), name it, and set a slug
- Open the Tools tab and assign the tools this gateway should serve
- Open the Authentication tab and add an auth method before going live

The gateway’s Integrations tab gives you the live endpoint plus discovery URLs to paste into any client:

Protocols
One set of tools, four ways to reach them. Start with the protocol your client speaks:
| Protocol | Reach it with | Endpoint |
|---|---|---|
| MCP | Claude, Cursor, Windsurf, VS Code Copilot | api.almyty.com/{org}/{gateway} |
| A2A | Google A2A agents, via the Agent Card | api.almyty.com/a2a/{org}/{gateway} |
| UTCP | Any HTTP client or agent framework | api.almyty.com/utcp/{org}/{gateway} |
| Agent Skills | 30+ coding agents, via the Skills CLI | Installed as SKILL.md files |
Agents you build on almyty also get an OpenAI-compatible endpoint, so anything that talks to the OpenAI API can call an agent without code changes.
Control what each gateway exposes
- Authentication: default-deny on every gateway. Add API key, bearer, basic, JWT, OAuth 2.1, and manage keys per client.
- Tool scoping: choose exactly which tools a gateway serves, with per-gateway rate limits and security policies on each tool.