> ## Documentation Index > Fetch the complete documentation index at: https://www.revve.ai/docs/llms.txt > Use this file to discover all available pages before exploring further. # Ask Revve: Your In-App Copilot > For every Revve user: chat with the built-in copilot to analyze calls, create agents and campaigns, and propose prompt changes — with a review-and-apply safety net for app owners. **Ask Revve** is the AI copilot built into the Revve dashboard. Ask it questions in plain language — "why did this call fail?", "show calls with negative feedback" — and it searches your conversations, calls, and metrics to answer. It can also do work for you: create a voice agent, set up evaluation metrics, draft a campaign, or propose edits to an agent's prompt that you review before anything changes. It never publishes anything. Changes to your agents arrive as proposals, apply only to a draft, and going live remains a human decision. ## Where to find it Ask Revve is one click away on every dashboard page, via the pill dock at the bottom right: * The orange **Ask Revve** button (sparkles icon) starts a fresh conversation. * Up to three pills show your most recent conversations — click one to pick it back up. * The **Chat history** clock icon opens your full conversation history. * The chevron (**Collapse dock** / **Expand dock**) tucks the dock away when you need the space. The Revve dashboard home: the Ask Revve composer with three suggestion chips beneath the copilot illustration, and the Ask Revve dock with recent conversation tabs at the bottom right The dashboard home page *is* the copilot: landing on the dashboard shows a centered composer with three suggestion chips — **Summarize recent voice calls**, **Show calls with negative feedback**, and **Help me improve a agent prompt** — so you can start with a question instead of hunting through menus. Once a chat is open, it runs in a floating window at the bottom right. Use **Expand** to take it full-surface and **Minimize** to shrink it back. The header shows the conversation title (or "New chat") and a status dot that pulses while the copilot is responding. ## Conversations and history Each chat is a thread with its own context. **Ask Revve** always starts a new one; the dock pills are your three most recent. For anything older, open **Chat history** — it has a **Search chat history** box and groups threads by Today, Yesterday, Last week, 2 weeks ago, Last month, and Older. Threads title themselves automatically after the first exchange, so you don't need to name anything. Keep one thread per task rather than one giant thread. "Improve the greeting for my support bot" and "why did enrollment drop last week?" are separate investigations — separate threads keep the copilot's context focused and make the history searchable. ## It knows what you're looking at On a chat agent, voice agent, or call-detail page, Ask Revve attaches the thing you're viewing as a context chip in the composer (for example **This call**). That means "improve this bot's greeting" or "why did this call fail?" just works — you don't have to name the agent or paste a call ID. Click the chip's dismiss control if you want a question answered without that context. Other pages attach no context, so name what you're asking about. ## What it can do Ask Revve's capabilities come in three tiers, ordered by how much they change: | Tier | Examples | Review step? | | ------------------------- | ---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | ------------------------------------------------------- | | **Answer and analyze** | Search conversations, calls, and metrics; summarize recent calls; surface negative feedback | None — read-only | | **Act directly** | Create a voice agent; create [evaluation metrics](/docs/evaluations/evaluation-metrics) and attach them to an agent; create custom contact fields; create a [campaign](/docs/campaigns/creating-a-campaign) and attach an agent; save skills | None — happens immediately (requires **Manage Agents**) | | **Propose agent changes** | Edit a chat or voice agent's prompt or conversation flow | Yes — you review a diff and apply per change | Direct actions create *new* things — a new agent, a new metric, a new campaign. One caveat to know: a copilot-created campaign lands **active** (though empty — no contacts are enrolled until you add them), not as a draft. Anything that would modify an existing agent's prompt or flow goes through the proposal flow below. ## Proposals: review before anything changes When you ask Ask Revve to change an agent — "make the greeting warmer", "add an objection-handling step to the flow" — it doesn't edit the agent. It replies with a **proposal card** titled "Proposed N change(s)" for that agent, with a per-change **diff** showing exactly what would change. From the card you can: * **Apply** an individual change, or use **Apply all** / **Apply selected (n)** for several at once. * **Discard all** / **Discard selected (n)** to reject changes you don't want. * Edit the proposed text inline before applying — supported for the system/global prompt and for node instructions — when the suggestion is close but not quite right. Applying opens a confirmation dialog — **Apply to draft?** — that spells out the contract: the changes are applied to the agent's *draft* only, and won't go live until you publish. After you confirm, the card shows an **Applied** badge and a **View draft** link that takes you to the agent's General tab to inspect the result. The safety story, stated plainly: * Proposals write to the agent's **draft**, never to the live published version. * **Publishing stays a separate, human step** — the same [draft/publish workflow](/docs/getting-started/publishing-and-versioning) as any manual edit, including [publish approval](/docs/governance-access/publish-approval-workflow) if your workspace requires review. * Proposing and applying agent changes both require the **Manage Agents** permission — without it, the copilot declines to propose changes at all (and conversations are private to each user, so there is no hand-off-to-a-teammate path). Drafts are single-editor. If a teammate holds the agent's edit lock, applying fails with `Draft is locked by {name}. Try again later.` — coordinate with them or wait for the lock to release. ## Permissions at a glance | Action | Who can do it | | ------------------------------------------------------------------ | -------------------------------------------------- | | Chat with Ask Revve, run analyses | Any team member | | Direct actions (create agents, metrics, campaigns) | **Manage Agents** | | Apply a proposal to an agent's draft | **Manage Agents** permission | | Create custom contact fields or team-scoped skills via the copilot | **Manage Settings** permission | | Publish an agent draft | Unchanged — your normal publish flow and approvals | See [Roles and Permissions](/docs/governance-access/roles-and-permissions) for how these permissions map to roles. ## Teach it your playbooks with skills The **Skill** button (book icon) in the composer lets you invoke a specific skill — a reusable, named instruction module — for a message, and the **Manage skills** link opens the Skills page where you author your own. Skills are how you turn a one-off prompt like "review this week's calls the way our team does" into a procedure every teammate's copilot follows the same way. See [Skills](/docs/ask-revve/skills) for scopes, authoring, and the built-in skills that ship with Revve. ## What's Next * [Skills](/docs/ask-revve/skills) — encode your team's workflows so Ask Revve follows them consistently. * [Publishing and Versioning](/docs/getting-started/publishing-and-versioning) — the draft/publish model that keeps applied proposals from going live on their own.