---
title: "WorkbenchLM"
url: "https://docs.mekodata.ai/guides/workbench-lm/"
---

# WorkbenchLM

Chat with your datapack in the Meko portal without installing an MCP client

**WorkbenchLM** is Meko's built-in chat experience in the [Meko portal](https://cloud.mekodata.ai/). Use it to try memory, knowledge, and traces without configuring Cursor, Claude, or another external MCP client.

Free tier

Free accounts are limited to **1,000 conversations** and **10M WorkbenchLM tokens**. When a cap is reached, the chat composer is disabled.

When the composer is disabled, the banner shows the conversations or WorkbenchLM tokens you have used. On Free, choose **Upgrade to Standard**. On Standard, choose **Talk to our team**. See [Reach a tier limit](/architecture/core-concepts/datapacks/#reach-a-tier-limit).

## When to use WorkbenchLM

- First session: Onboarding offers **Try the WorkbenchLM** so you can explore a sample knowledge base and see memories form in real time.
- Quick experiments: Upload a document, ask questions, and inspect traces and memories from the same UI.
- No local install: WorkbenchLM runs entirely in the portal; you do not need to set up an MCP server with an MCP config file or run the installer.

## Use WorkbenchLM

1. Sign in to your account.
   
   The first time you sign in, Meko displays the **Welcome to Meko** page.
2. Click **Try the WorkbenchLM**.

If it is not your first time signing in, do the following:

1. Sign in to the Meko portal.
2. Select **WorkbenchLM** in the left sidebar.

## What you can do

- Pick a datapack: Select a datapack from the menu in the header. The list includes datapacks you own and datapacks shared with you as **Maintainer** or **Contributor**. (Viewer-only datapacks are not listed, because chat writes memories.) You can also search for datapacks by name or role. If you don't have any datapacks, the picker says **No datapacks available**. A search with no hits says **No matching datapacks**.
- Open WorkbenchLM from a datapack: On any datapack page, open the actions menu (the gear next to **Share**) and choose **Open WorkbenchLM**. That item appears only when you can chat with the datapack, and WorkbenchLM opens with that datapack selected.
- Knowledge: Upload or manage knowledge-base documents in the left panel. Chat is available after a document has finished indexing.
- Chat: Send messages in the center panel. WorkbenchLM can call Meko tools while it answers. A read runs immediately. Saving a memory does too. Deleting a memory, promoting a memory, or updating the datapack waits until you choose **Allow**. **Deny** leaves it unchanged.
- Live observation: The right panel shows memories, shared knowledge, and conversations, updated as you chat.
- Traces: Expand **Decision trace** above each reply, or click **View execution trace** under it to view decision traces. See [Observability](/guides/observability/).

## WorkbenchLM vs Choose your tool (via external MCP clients)

|                | WorkbenchLM                   | Choose your tool                                              |
|----------------|-------------------------------|---------------------------------------------------------------|
| Setup          | None in the portal            | Shared MCP URL (`https://mcp.mekodata.ai/mcp`) + user API key |
| Authentication | Your portal session           | Long-lived `mko_tkn_…` key from **Settings>API Keys**      |
| Best for       | Trying Meko, demos, light Q&A | IDE-integrated development                                    |

External clients authenticate with an [API key](/integrations/#api-key). WorkbenchLM uses your signed-in session and server-side credentials for the selected datapack — you don't need to provide an API key to use WorkbenchLM.

## Next steps

[**Connect a coding agent**  
\
Install Meko MCP on Cursor, Claude Code, or Claude Desktop.](/integrations/)

[**Work with memory**  
\
Add, search, and manage agent memories.](/guides/working-with-memory/)