Relace supports a selection of predefined agents that can be run on any Relace Repo. Our agents are task-specific and run on our own code execution sandboxes.

​
Why Use Relace Agents?

There are two main reasons to use Relace Agents:
  • You need a coding agent but don’t want to manage infrastructure, define toolsets, or maintain evals.
  • You have your own AI codegen system and want to use our utility agents to quickly retrieve relevant files, refactor codebases, or fix bugs.

​
Coding Agents as a Service

We’ve provided models for 40+ codegen companies and have seen dozens of different implementations of coding agents. A team of three people would need at least 1-2 months to set up an agent and would be occupied full-time maintaining it. If you don’t want to sink resources into this, you can incorporate codegen into your product with the Relace agent in less than a week. Relace agents use heavily tested toolsets and latency-optimized infrastructure, which are updated regularly with best practices. Security can also be challenging to enforce with systems that execute arbitrary code — there have already been a handful of sever vulnerabilities discovered in high profile codegen companies. All code executed by Relace agents are sandboxed to counter any possible code injection attacks. Focus on building your product, and leave the messy configuration details to us.

​
Task-specific Agents

There are a handful of critical workflows within code generation, which we’ve trained small, fast agents to perform.
  • Small agentic search for fast & flexible retrieval
  • Refactoring large files within codebases
  • Static analysis and bug fixing
These models are still in private beta, but they will be released soon. Reach out to us for early access.

​
Data Analysis Agent

A common use case for codegen is generating custom data visualizations and dashboards for users. We’ve built a specialized agent to facilitate this use case. The Relace analysis agent transforms structured data into interactive web dashboards using a Vite.js template with React and Tailwind CSS. You start by creating a Relace Repo from our base template, then upload your CSV or JSON files to the public folder where the agent can access them. The agent accepts a custom query describing what analysis you want and an optional system prompt to guide its behavior. It processes your data to identify patterns, calculate statistics, and generate insights, then builds interactive visualizations using Chart.js or similar libraries. The entire analysis is packaged into a comprehensive dashboard with multiple views. Once execution completes, your dashboard is automatically deployed and accessible via a public URL that follows the pattern https://{repo_id}.dev-server.relace.run. The generated dashboard can be embedded as an iframe in other applications or viewed directly in the browser. You can iterate on your analysis by sending follow-up requests to the same repository. The agent will enhance the existing dashboard with additional features, views, or export capabilities based on your new requirements, building on the previous work rather than starting from scratch.

​
Next Steps

For more details on running the Relace anlysis agent, see our step-by-step guide or explore the API reference.