> For the complete documentation index, see [llms.txt](https://docs.finchagentic.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.finchagentic.com/agents.md).

# Agents

Persistent agents are named workstreams that **keep state** — not one-shot function calls.

## Mental model

| Concept       | Meaning                                          |
| ------------- | ------------------------------------------------ |
| Spawn         | Create a named agent with a goal                 |
| Recall        | Load latest state weeks later                    |
| Update        | Progress / notes                                 |
| Schedule      | Recurring autonomous runs (needs confirmation)   |
| Ledger / runs | Audit of what ran, cost, status                  |
| Identity      | Optional Base address (custodial — see security) |

## Typical prompts

```
spawn an agent called market-researcher
goal: track Base protocols weekly

what is agent market-researcher working on?
update agent market-researcher progress: "covered Aerodrome and Morpho"
```

## Tool surface (MCP)

| Tool                                  | Role                        |
| ------------------------------------- | --------------------------- |
| `agent_spawn`                         | Create tracker / agent note |
| `agent_recall`                        | Resume state                |
| `agent_update`                        | Progress                    |
| `agent_schedule` / `agent_unschedule` | Recurring runs              |
| `agent_pause` / `agent_resume`        | Control                     |
| `agent_identity`                      | Base identity address       |
| `agent_ledger` / `agent_runs`         | Audit                       |
| `list_agents`                         | List                        |
| `hire_agent`                          | One-shot specialist run     |

> Product also has richer “create agent” paths in the app (Identity + Prompt + Skills). MCP `agent_spawn` is the multi-session tracker path; don’t confuse with one-shot `hire_agent`.

## Confirmation required

Before `agent_schedule`:

1. Name + goal
2. Cadence (daily / weekly / cron)
3. Expected LLM cost per run
4. Whether vault writes are allowed
5. Explicit user yes

## Identity custody (critical)

`agent_identity` returns a **backend-controlled** Base address. The private key is not the user’s browser wallet.

* Do **not** casually tell users to fund that address
* User login wallet ≠ agent execution identity
* See [security.md](/security.md) boundary 8

## Related

* [Workflows](/workflows.md) for automations/monitors without a full agent persona
* [Tools: agents](/tools/agents.md)


---

# Agent Instructions
This documentation is published with GitBook. GitBook is the documentation platform designed so that both humans and AI agents can read, navigate, and reason over technical content effectively. Learn more at gitbook.com.

## Querying This Documentation
If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter, and the optional `goal` query parameter:

```
GET https://docs.finchagentic.com/agents.md?ask=<question>&goal=<endgoal>
```

`ask` is the immediate question: it should be specific, self-contained, and written in natural language.
`goal` is optional and describes the broader end goal you are ultimately trying to accomplish on behalf of the user. GitBook uses it to tailor the answer towards what is most useful for that goal.

The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.