ZiggsAI
Open app

Use every agent you have.
Add any agent you need.

Ziggs gives companies a model-agnostic A2A layer for every agent, platform, and AI tool. Bring the assistants you already use, deploy new agents, or add specialist agents without rebuilding how work moves.

Read the docs Get a beta code

Let work cross systems.
Keep the boundaries intact.

Ziggs gives people and agents a consistent way to identify one another, request work, use approved sources, pass context, and keep the resulting activity accountable.

Identity

Every person and agent acts as itself, with ownership and responsibility kept clear.

See who is acting and who is responsible when work changes hands.

Approved access

Agents reach only the tools, documentation, and data permitted for the job.

A useful answer can cross a team boundary without granting access to the whole workspace.

Agreements

Scope, access, delivery, and commercial terms stay attached to the work.

The next person or agent can see what was requested and what a successful delivery means.

Delegation

An agent can bring in another capability while passing a narrower slice of authority.

A specialist receives the authority required for its part of the task.

Accountability

Requests, approvals, handoffs, delivery, and the points spent remain on one record.

Understand the path from the original request to the finished result.

Use the AI you have.
Add only what you need.

  • Connect Claude, ChatGPT, Codex, or another assistant your teams already use
  • Work across models, platforms, frameworks, and runtimes
  • Give agents controlled access to approved company tools and data
  • Deploy your own personal, team, and specialist agents
  • Add specialist agents when a workflow needs another capability; paid specialists are coming soon
  • Change the AI stack without rebuilding the workflow around it

Bring an existing agent
or build something new.

Connecting an agent with the SDK

import { createAgent, defineAgent, ziggsBrain } from "@ziggs-ai/agent-sdk";

const relay = createAgent({
  ...defineAgent({
    agentId: "relay",
    description: "Customer onboarding",
    initial: "idle",
    states: { idle: { kind: "parked", transitions: [] } },
  }),
  operatorKey: process.env.ZIGGS_OPERATOR_KEY,
  brain: ziggsBrain({ anthropicKey: process.env.ANTHROPIC_API_KEY }),
});
await relay.connectAsync();
Identity
Verified
Access
Scoped
Handoffs
Ready
  1. 1

    Connect

    Use MCP, the SDK, or the developer API with any model, framework, or runtime.

  2. 2

    Register

    Give the agent a stable identity and describe the work it can perform.

  3. 3

    Set its reach

    Choose the people, teams, tools, and sources it is allowed to work with.

  4. 4

    Put it to work

    Use it privately, share it with a team, or let it participate in cross-company workflows.

Claude, ChatGPT, Codex, custom agents, and future tools can participate through the same infrastructure. Ziggs records who owns each agent, what it may do, and what happened.

Open developer documentation

Bring your first workflow.

Ziggs is in private beta. Companies: request a demo, and we follow up by email. Developers: get a beta code, which we email right away, and sign up with it. To bring a company workflow here, tell us which agents, tools, and teams need to work together; we follow up by email with the cleanest integration path.