Automate with agents
Put an AI teammate on the work — in a conversation or on demand, bounded by a role you control.
Agents are AI teammates that work inside your Augno account. An agent has a name, a set of standing instructions, a role that bounds what it is allowed to see and change, and a set of tools it can call. When something starts an agent — a person in a chat, or an explicit trigger — Augno creates a run, and the agent works the problem step by step until it answers, finishes, or stops to ask you something.
Agents are not a separate product bolted onto Augno. They act through the same API your team and your integrations use, authorized against the same roles and permissions, and everything they do is recorded on the run.
What you need to use agents
Agents are a paid-plan feature. Accounts on the free plan cannot use them: the API rejects every agent operation with an authorization error, and the dashboard redirects free-plan users away from the AI Agents area. See Plans & billing to upgrade.
Within a paid account, the AI Agents area is visible to admins and to anyone whose role grants read access to agents or to agent runs. Creating and editing agents requires create/update permission on agents. Starting a run requires create permission on agent runs; resuming, retrying, and cancelling a run require update permission on agent runs.
Where agents live
Everything is under AI Agents in the dashboard:
| Area | What it is |
|---|---|
| Overview | Summary counts and the most recent runs |
| Agents | The agent list — filter by status, trigger type, and system vs. custom |
| Agent Runs | Every run for the account, with the full step-by-step timeline for each |
| Memories | Notes saved for agents to recall on later runs |
Memories belong to the account rather than to one agent, so a note you save is recalled by every agent in the account — see Memories.
The ways you meet an agent
In a conversation. An agent can be a participant in a chat thread. Someone mentions it, it reads the thread, does the work, and posts its reply back into the same conversation. Its reasoning streams live while it works, and you can reply to keep the same run going rather than starting a fresh one. See Agents in chat.
By hand. Any agent can be started explicitly at any time. A run started that way is always recorded as a manual run. See Agents for how triggers are configured.
What an agent can actually do
Two kinds of tools:
- Built-in tools — create an artifact, read Augno's documentation, fetch a public URL, send an email reply, or draft a reply for a human to approve.
- Endpoint tools — Augno's own API operations, exposed to the agent as tools. There are over 200 of them, and an agent gets exactly the ones you grant it, and no more.
Every tool call is still authorized against the agent's role, so a grant can only ever narrow what the role already allows — never widen it. And anything you want a human to sign off on can be gated so the run pauses and waits for approval. See Tools & approvals.
Agents
Define an agent: instructions, trigger, intelligence tier, and role.
Runs
Follow a run, read its timeline, and understand the safety rails.
Tools & approvals
Grant tools, gate the risky ones, and approve or deny a paused run.
Memories
Durable notes every agent in the account recalls on later runs.
Agents in chat
Put an agent in a conversation and work with it in the thread.
Next: Agents