Skip to content

How-To Guides

These guides help you get the most out of ado depending on what you want to do. Choose a path that matches your role, or read on to see how ado creates a powerful partnership with coding agents.

Choose Your Path

  • Researchers & Benchmarkers — You want to run experiments, collect data, and analyse results, using the tools already available in ado, or building new ones as needed.
  • 🔌 Plugin Developers — You want to write new experiments, analysis tools or sampling methods, extending ado's capabilities for everyone.
  • Core Developers — You want to contribute to the ado framework itself

ado and Coding Agents

ado is designed from the ground up to partner with coding agents, creating a powerful automated research assistant. This isn't just about agent skills; ado's core design allows an agent to reason about and execute a complete research workflow.

Step Capability Benefit for Agent-Driven Research
1. Discover Self-Describing Experiments and Operators Before acting, an agent must understand what's possible. ado enables agents to discover exactly what capabilities are available, and what's required to use them, in a structured manner rather than parsing code.
2. Model Clear Separation of Concerns Once possibilities are known, the agent structures the problem. ado provides clear separation between the what, the how, and the action, allowing the agent to reason about each part of the problem independently.
3. Act & Verify Pre-Run Validation With a model in mind, the agent can safely execute. Using ado template and ado create --dry-run, it can create a tight generate → validate → fix → run loop.
4. Analyze & Refine Structured & Queryable Data All data and metadata created via ado is stored in a structured database. This allows the agent to analyze, compare, link, and synthesize data to decide on the next course of action.

Together, these properties enable a closed research loop: an agent can describe a problem, run experiments, read the measurements, and refine its approach, all while operating at a high level of abstraction — manipulating spaces and operators rather than writing bespoke glue code.

See Core Concepts and The ado CLI for more.