Developers

Publish agents through the interfaces your product already uses.

Turn a configured agent into a verified REST endpoint or an MCP server, then follow every execution through its operational trace.

Sample published agentIllustrative

Example endpoint · POST

/api/v1/public/v1/agt_x7k2m9p4

Interface

REST · MCP

Credential

zak_live_••••

01 · Run an agent

Make the first call without a client library.

Every published agent has a public REST endpoint. Use any HTTP client your stack already trusts.

curl -X POST 'https://api.zilionix.com/api/v1/public/v1/agt_x7k2m9p4' \
  -H 'X-API-Key: zak_live_your_key' \
  -H 'Content-Type: application/json' \
  -d '{"message":"Summarize today’s priority tickets"}'

02 · Publish

A published agent carries its own delivery path.

Keep configuration, access, invocation, and operational evidence connected from the first publish through every execution.

01

Configure and publish an agent

Set the agent’s instructions, tools, model settings, and access controls, then publish it.

02

Receive a publish ID and API key

Use the publish_id and the one-time zak_live API key created for that published agent.

03

Choose the delivery interface

Call the public REST endpoint directly or enable the agent’s MCP server for compatible clients.

04

Inspect the execution receipt

Review execution, tool, token, cost, and latency traces after each run.

03 · Interfaces

Integrate the agent without losing the operational context.

REST execution and streaming

Invoke published agents over REST and stream responses when the experience calls for it.

Per-agent MCP server

Expose an individual agent as an MCP server for MCP-compatible clients.

Event-driven invocation

Use webhooks along with scheduled and event triggers to start work on the right signal.

OpenAPI tool import

Bring REST APIs into an agent’s toolset from an OpenAPI description.

Scoped credentials

Keep credentials bound to the agents and tools that need them.

Trace and cost telemetry

Inspect execution traces alongside tool, token, cost, and latency data.

Model and connector surface

OpenAI, Anthropic, AWS Bedrock, Azure OpenAI, and DeepSeek, with 100+ models available through provider discovery and 114 MCP connector presets plus native actions and OpenAPI import.

OpenAI
Anthropic
AWS Bedrock
Azure OpenAI
DeepSeek
Slack
GitHub
Google Calendar
Notion

Deployment options

Deploy in Zilionix cloud, a customer VPC, with Docker and Terraform in self-hosted infrastructure, or on premises.

Zilionix cloud

Customer VPC

Docker and Terraform

On premises

Documentation

Keep the endpoint contract close to your implementation.

Start with the verified request shape, then follow the publishing flow for the agent interface your product needs.

Give every agent a path into the products people use.

Configure the work, publish the interface, and retain the trace behind every execution.