Console Overview
The console is where your team builds, tests, launches and oversees AI agents, all through the web browser and without writing code. Its menu on the left is organised into five sections that follow the life of an agent: Build, Resources, Evaluate, Operate and Govern. This page tells you which page to open for each job.
Work in the console is organised by project. A project groups related agents, data connections and channels. Use the project switcher at the top of the sidebar to change project, and the menu follows your choice. Items marked Soon are visible but not available yet.
Find the right page by task
| I want to | Go to |
|---|---|
| Create my first agent from my procedures | Build, Agent Builder |
| See and edit all agents in a project | Build, Agents |
| Check activity and cost at a glance | Build, Overview |
| Connect a database or system | Resources, Datasources |
| See what abilities exist and who can use them | Resources, Tools |
| Try the agent as a customer would | Operate, Playground |
| Put the agent on my website or WhatsApp | Operate, Deployments |
| Read a specific conversation | Operate, Sessions |
| See what my team changed | Operate, Audit Log |
| Review the rules that protect data | Govern, Guardrails |
Build
Where you create and shape agents.
| Page | What you use it for |
|---|---|
| Agents | The list of agents in the project, where you open one to review or change how it behaves. Problems are flagged as you edit, and saving is blocked while something is invalid. |
| Overview | A snapshot of conversations, messages, cost estimate and handoffs waiting, plus a warning if any agent is not ready to deploy. |
| Agent Builder | A guided flow that reads your procedures and connects to your data to draft an agent for you to review. See Agent Builder. |
| Agent Graph | A picture of one agent and the abilities it can use. |
| Workflows | Soon. |
Resources
What your agents connect to.
| Page | What you use it for |
|---|---|
| Tools | A read-only list of every ability across your agents, what it does, and which roles can use it. See Agent abilities. |
| Datasources | Where you add and manage the credentials that let an agent reach your systems. Credentials are stored securely. See Connect your data. |
| Integrations | The health of connected services and speech providers, and the switch for voice. |
| Knowledge Bases, Prompt Library, External Agents | Soon. |
Evaluate
Making sure the agent behaves before and after changes.
| Page | What you use it for |
|---|---|
| Evals | Run saved test questions against an agent and see which passed. See Test your agent. |
| Experiments | Soon. |
Operate
Running agents day to day.
| Page | What you use it for |
|---|---|
| Playground | Chat with the agent as any role, with voice, and inspect what it did. The best place to try things safely. |
| Sessions | Every conversation, with a replay and timeline. See Audit trail. |
| Audit Log | What your team changed, and a summary of refusals and errors. |
| Health | A score and deploy-ready verdict for each agent. See Monitor your agent. |
| Deployments | Publish changes and manage channels such as the website widget and access keys. See Website widget and Messaging channels. |
| Transfer Sessions | Conversations the agent passed to a person. See Human handoff. |
| Inbox | Soon. |
Govern
| Page | What you use it for |
|---|---|
| Guardrails | A read-only view of the safety rules for the selected agent: who can use which abilities, and which records each role can see. The rules are enforced by the platform itself. See How Achilles keeps agents safe. |
The Arch assistant
A panel on the right of the agent editor reviews the agent you are editing, points out problems and suggests fixes. It proposes a change and shows what would differ. Nothing is saved until you accept it.
Signing in and access
You sign in with your own account, and the console only shows workspaces you belong to. Sessions last 12 hours. Everything your team does to agents is recorded in the Audit Log. See Users and teams.
Suggested first week
- Use Agent Builder to draft your first agent.
- Add your data in Datasources.
- Try it in the Playground under each role.
- Run Evals and check Health.
- Deploy from Deployments and add the widget.
- Start your review routine with Sessions.
If you are new to the terms, read Key terms or start with the Quickstart.