Labs
Use Labs to explore core Meko features with a live demo application connected to a real datapack. Each lab walks a stage of the agentic development lifecycle: you build the context, an agent uses it, and you see exactly what it changed. Every lab ends with a path back to your own repo. Nothing to install.
Labs sit in that lifecycle — Spec, Build, Review, Fix & merge — with Review live today.
To run a lab:
- Sign up.
- Log in.
- In the portal sidebar, open Labs, choose a scenario, and click Launch lab.
- Follow the on-screen instructions.
Labs
The following lab is available (with more in development).
Review a PR with AgentK
Learn how an agent reviews a pull request against your org's coding standards, live. You start with a sample project that has no standards yet, build that context in a datapack, then watch AgentK use it.
The lab takes about five minutes across five chapters:
- Give your agent context. Watch AgentK review a pull request with no standards, then build the datapack it was missing: a standards document, a memory, and a conversation.
- Share it with AgentK. Grant the agent read access to that datapack and choose the model and sources for the run.
- Read the review. Run the review and compare it with the review that had no context.
- Close the loop. Save the findings into the datapack, then promote one memory so other agents on the team can read it.
- See it in production. Look at the same loop running on Yugabyte's repositories, then optionally point AgentK at a pull request of your own.
AgentK runs on AWS Bedrock AgentCore.
Limitations
- Creating a datapack during a lab counts against your plan's datapack quota. Free accounts can create up to 10 datapacks. See Meko pricing.
- Free accounts use the shared, predeployed AgentK. Deploying a dedicated instance requires a paid plan.
- Pointing AgentK at your own pull request is optional. Your GitHub personal access token stays in the browser; Meko never receives it. The PR title and full diff are then sent to Meko and the review agent — on the Free tier, the shared predeployed instance.