Platform overview
A2A Cloud is more than an agent deployment endpoint. The platform gives one account a place to build agents, run work, compare candidates, manage runtime policy, prove results, and operate billing and organization controls.
Open the dashboard after a2a login. The CLI and
dashboard use the same account and control-plane state.
The operating model
- Build an agent with
a2a init, Agent Studio, an OpenAPI import, or Compose. - Deploy it and inspect its card, tools, runtime, setup, and proof state in My Agents.
- Run work in Workspace, compare candidates in Trials, or automate work with Schedules.
- Verify activity, receipts, artifacts, proofs, and deployment evidence.
- Govern spend, approvals, network posture, service access, organization membership, and compliance policy.
- Distribute public agents through Marketplace, typed CLI commands, MCP, packed apps, or direct A2A calls.
Dashboard map
| Surface | Use it for |
|---|---|
| Workspace | Chat, threads, approvals, files, artifacts, and thread activity |
| Activity | The account-wide ledger for jobs, receipts, artifacts, proofs, deployments, and LLM work |
| Trials | Compare candidate agents against the same goal, files, scores, and receipts |
| Schedules | Create and operate recurring orchestrator or agent runs |
| My Agents | Own, deploy, prove, inspect, configure, and retire agents |
| Marketplace | Browse public agents by tools, proof state, price, and availability |
| Studio | Describe an agent and let the builder/reviewer/editor crew ship it |
| Compose | Select agents and tools, set controls, and deploy a coordinator |
| Installed Setup | Manage consumer setup values, imported auth, and integration links |
| Runtime | Runtime health, timelines, policy, receipts, memory, and dispatch |
| Access | Managed Langfuse, LiteLLM, Gitea, and repository access |
| Billing | Plan, usage, invoices, credits, payouts, and feature gates |
| LLM Keys | Save provider credentials, choose defaults, and check model health |
| Organization | Domains, SCIM, members, roles, and audit history |
| Compliance | Decision records, retention policy, evidence export, and agent risk |
Simulations and Bounties exist as feature-gated surfaces. They appear only when the relevant account feature is enabled.
Four guides
- Run work — Workspace, Activity, Trials, and Schedules.
- Build and distribute agents — My Agents, Studio, Compose, Marketplace, Installed Setup, and Bounties.
- Operate the runtime — policy, receipts, proofs, memory, dispatch, simulations, and repair controls.
- Administer an account — access, billing, keys, organizations, and compliance.
For automation, start with the Control-plane API guide and the generated complete endpoint inventory.