Build with Inventor

A copilot agent inside the Relevance AI UI that builds, debugs, and iterates on agents, tools, workforces, and triggers from a natural-language description.

Inventor is a pilot for Enterprise plans only. To request access, reach out to your account team. On another plan? See how to invent an agent or invent a tool.

What is Inventor

Inventor is a copilot agent inside the Relevance AI UI. It uses the Relevance AI MCP under the hood to take real actions on your project, so from a single chat surface you can build, debug, and iterate without leaving the platform.

What you can do

Build from a description

Create new agents, tools, workforces, and triggers from a natural-language description.

Debug failing agents

Inventor reads run history, identifies the failing step, explains what went wrong, and proposes a fix.

Iterate on existing assets

Change a prompt, swap a model, restructure a workforce, or regenerate a tool’s schema.

Restore a version

Roll an existing agent, tool, or workforce back to an earlier version, with a required confirmation step.

Discover before you build

Find existing agents and tools before building duplicates.

Recommend the right abstraction

Get a recommendation on tool vs. agent vs. workforce based on your use case.

Build and improve evals

Diagnoses errored logs and tools, explains why a run failed, and proposes how to fix it — so you can build stronger evals.

Log bugs to engineering

When Inventor hits a platform issue it can’t fix, it logs a bug to our engineering team on your behalf.

Inventor can log bugs on your behalf. These go to our engineering team, not support, so you won’t receive updates on their progress.

Where to find Inventor

Inventor is available from several places, so you can reach it wherever you’re working.

Home page

Inventor has its own item in the left sidebar that opens a full home page — a central hub for all your Inventor conversations. The same Inventor is available as a persistent side panel from any builder page — tool, agent, workforce, knowledge, and home — so you can reach it without navigating to a specific route.

Start Inventor from the home page:

1. Open Home

   Click Home in the left sidebar.

2. Open Inventor

   Click the Inventor button in the page header.

While editing an agent

Inventor sits at the top of the agent builder, above Prompt, Tools, Knowledge, and Triggers. Open it to iterate on the agent you’re editing without leaving the page.

Start Inventor while editing an agent:

1. Open the Agents page

   Click Agents in the left sidebar.

2. Open or create an agent

   Open an existing agent or create a new one to enter the agent editor.

3. Open Inventor

   Click the Inventor button in the left sidebar of the agent editor.

When creating an agent

Start Inventor from the Agents page:

1. Open the Agents page

   Click Agents in the left sidebar.

2. Build with Inventor

   Click the Build with Inventor button in the header.

When creating a tool

Start Inventor from the Tools page:

1. Open the Tools page

   Click Tools in the left sidebar.

2. Build with Inventor

   Click the Build with Inventor button in the header.

You can also choose Invent when you create a tool to generate a first version from a description.

How to use Inventor

Inventor is expensive to run and consumes significant credits. Use it deliberately and keep an eye on your usage.

1. Open Inventor

   From the Relevance AI home dashboard, click Inventor.

2. Describe what you want

   Type a natural-language description of what you want to build, debug, or iterate on — agents, tools, workforces, and triggers. Be specific about the outcome you’re after.

3. Review and iterate

   Inventor generates a first version or applies your edit. Review it, then refine by describing the changes you want.

Credit usage

The Inventor panel shows a real-time credit counter — “X credits used” with a coins icon — that updates with each interaction, such as generating an agent, requesting an adjustment, or asking a follow-up question.

The counter lives in the Inventor panel and tracks only the credits used in your current session.

Inventor features

Auto-accept

By default, the changes Inventor makes land in draft so you can review and approve them before they go live. Turn on auto-accept to let Inventor apply its changes as it works, without stopping for your approval. Version restores are the exception — Inventor always asks you to confirm before restoring. Keep it off when you want a checkpoint on each change, and turn it on when you trust Inventor to iterate quickly.

Model selection

The model dropdown sets which LLM powers your Invent session. Choose from the configurable list; there’s a default if you don’t change it.

About you

About You tells Inventor who you are so it tailors every response to your role and work from the first message — no re-explaining your context each session. It’s descriptive only: it shapes how Inventor talks to you and what it assumes about your work, but it doesn’t change Inventor’s role or hardcode any tool configuration.

What it changes:

Tone and language

Simpler explanations for non-technical users, more precise technical language for engineers.

Domain assumptions

If you’re in sales, it won’t default to engineering-flavored tool designs.

Personalization

Addresses you by first name at natural moments like greetings, not robotically in every message.

To set it up:

1. Open Inventor

   Open the Inventor side panel or home page.

2. Click Settings

   The settings pop up — enter your details under About You.

Conversations

Inventor saves each conversation as a session automatically. Opening the panel starts a new chat, and your past sessions stay available from the Chats dropdown in the panel header — browse or search them, then select one to resume where you left off.

Closing the panel or refreshing the page doesn’t delete your conversation history — sessions persist and can be resumed at any time.

Feedback

Inventor has built-in feedback tools that help improve its output over time. Each AI message in the panel has thumbs up and thumbs down buttons.

Thumbs up

The response was accurate and helpful.

Thumbs down

The response was incorrect, unhelpful, or the generated agent or tool didn’t match what you described.

When you click thumbs down, a popup appears where you can submit a bug report directly to the Relevance AI team — without leaving the interface or sending a separate email. Describe what you asked Inventor to build, what it generated, what was wrong or missing, and what you expected instead.

Restore a version

Ask Inventor to roll an asset back to an earlier version — for example, “Restore the previous version of this tool”. Inventor copies that version into the current draft, leaving the live published version unchanged until you publish. This restores an earlier version of an existing asset; it doesn’t recover one that’s been deleted.

MCP vs Inventor

Inventor and the Relevance AI MCP are often compared, but they solve different problems. Inventor builds things inside Relevance AI from natural language. The MCP connects Relevance AI to the outside world — letting external clients call your tools, and letting your agents reach into external services.

Inventor

A copilot inside Relevance AI. Describe what you want and it builds or edits agents, tools, and workforces for you, landing changes in draft for review.

Reach for it when you want to build, edit, or debug something inside Relevance AI by describing the outcome.

Relevance AI MCP

A bridge to other tools. Call your existing Relevance AI tools from external clients like Cursor or Claude with the MCP Server, or let your agents call services like Notion and Linear with the MCP Client.

Reach for it when the tool already exists and you want to run it from Cursor or Claude, or let an agent call an outside service.

The two are complementary, not competing. Inventor uses the same MCP tool surface under the hood — so a common pattern is to build your agents and tools with Inventor, then call them from Cursor or Claude through the MCP in your day-to-day work.