> For the complete documentation index, see [llms.txt](https://docs.zerve.ai/zerve-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.zerve.ai/zerve-docs/agents-in-the-workspace/evolve-mode.md). # Evolve Mode Zerve's most capable agent mode, which plans, delegates to specialists, checks its work and learns as it goes. Evolve Mode is Zerve's most capable agent mode. Rather than writing all the code itself, the agent acts as a **researcher**, **domain expert** and **quality gatekeeper**: it works out what's needed, hands the work to specialist agents running in parallel, checks what they produce, and keeps going until the request is completely and correctly done. It's the same agent that powers every [superagent](/zerve-docs/superagents/superagents.md) worker. Learning how Evolve Mode works is the best way to understand what a superagent's team is doing. ## When to use it Use Evolve Mode for substantial work, such as: * Research that needs current sources, such as a literature or market review * Building and validating a model or signal * Data pipelines that fetch, clean and combine data from several places * Anything with several steps, or where getting it right matters For a quick question or a small change, [Ask or Build](/zerve-docs/agents-in-the-workspace/agent-modes.md) is often faster. ## How it works ### It takes the fastest correct route Evolve Mode doesn't delegate everything. For each request, it picks the quickest way to get it right: | Request | What it does | | ----------------------------------- | ---------------------------------------------------------------------------------------------------------------------------------- | | **A question or quick explanation** | Answers directly | | **A small, self-contained change** | Makes the change itself, adding at most one new block | | **Anything bigger** | Delegates to specialists, for anything more than one block, or involving fetching data, modeling, pipelines or environment changes | ### It delegates to specialists For bigger work, Evolve Mode breaks the task down and hands parts of it to [specialists](/zerve-docs/agents-in-the-workspace/evolve-mode/specialists.md), such as coding, research and peer review agents. Several can work at once, for example multiple coding agents building different parts of a pipeline in parallel. You can see each specialist and its progress in the agent's plan. ### It checks before it answers Evolve Mode checks the notebook's actual state before acting, and doesn't claim something is done until it has verified it. It can ask a peer review specialist to check work before presenting it. It won't reply to you while any specialist is still working. It waits for them to finish, or stops them, first. ### It learns as it goes Evolve Mode keeps [Learnings](/zerve-docs/agents-in-the-workspace/learnings.md): persistent notes on the work's objectives, assumptions, chosen approach and research. They carry over between sessions, so the next time you return to the notebook, the agent picks up where it left off. ## Following its work As Evolve Mode works, you can follow: * **Its plan**, including each specialist it has delegated to and whether it's running, completed, failed or cancelled * **Its reasoning**, as a timeline of thinking, hypotheses, decisions, runs, observations and blockers * **The notebook**, as blocks are added and run ## Sending new instructions You can message the agent while it's working. Messages are queued and handled in order. If a new message changes what you asked for earlier, the **latest instruction wins**: earlier messages become context rather than extra tasks. ## Settings Evolve Mode has two settings: * **Credit cap**: the most a single run can spend. See [Credit caps](/zerve-docs/agents-in-the-workspace/evolve-mode/credit-caps.md). * **When it needs input**: whether the agent waits for your answer or carries on. See [When the agent needs input](/zerve-docs/agents-in-the-workspace/evolve-mode/when-the-agent-needs-input.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 by asking a question. Perform an HTTP GET request on the following URL with the `ask` and `goal` query parameters: ``` GET https://docs.zerve.ai/zerve-docs/agents-in-the-workspace/evolve-mode.md?ask=&goal= ``` `ask` is the immediate question: it should be specific, self-contained, and written in natural language. `goal` is what the user is ultimately trying to achieve, the reason they need the answer. Sharing it helps GitBook give you a better, more relevant answer. A goal is most helpful when it describes the outcome the user wants rather than restating the question. For example, with `ask=how do I create an API token`, a goal like `automate deployments from our CI pipeline` lets GitBook tailor the answer to that use case. 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.