Practical overview
When it fits
Use Eve when a task has a clear way to test or judge candidate results and repeated agent attempts could improve them.
Who it is for
- Developers and researchers exploring iterative, agent-generated solutions to evaluable coding or mathematical tasks.
- Users who can provide a working environment, define editable files and supply a scoring method.
Documented uses
- Run iterative experiments that improve coding-agent solutions and guidance.
- Compare candidate coding solutions using scripts, judge prompts or both.
- Refine agent instructions and skills alongside solver behavior.
Requirements & limits
The repository identifies Eve as open source under the Apache-2.0 license. Running agents requires separately configured provider credentials; Codex is the default backend.
- Tasks need a working environment, explicitly allowed edit locations and a scoring procedure.
- The framework evolves both solver candidates and agent guidance, using scores to inform later agent sampling.
- Agents run in separate workspaces, and designated modification scope is checked.
- Eve orchestrates third-party agents but does not provide AI service access or bypass provider authentication, rate limits or usage limits.
- The repository identifies Python 3.11 or newer and Hydra 1.3 in its project badges.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Set up Eve and run the documented Codex smoke test
Install the project dependencies, configure Codex authentication and run the repository's short circle-packing smoke-test configuration.
Before you start
- Python 3.11 or newer
- The Eve repository and its dependencies
- Codex installed with authentication configured
- For Codex 0.130.0 or newer, the documented hook-trust step
Steps
From the repository root, install dependencies with the command below.
uv syncInstall Codex and configure its authentication.
For Codex 0.130.0 or newer, run the command below, start
codex, trust all hooks through/hooks, then exit Codex.uv run python -m scaling_evolve.providers.agent.codex_hooksRun the command below.
uv run python -m scaling_evolve.algorithms.eve.runner --config-name=circle_packing.smoke
Expected outcome
The documented short circle-packing demo runs with headless Codex agents as a smoke test.
Guide limitations
- The smoke test is a setup check, not a guarantee of compatibility or successful results in other environments.
- Eve does not supply provider access or credentials; authentication, rate limits and usage limits remain with the provider.
Compare options for AI coding assistance
These projects document a common task and share a project role. This is a source-based comparison, not an execution benchmark or a guarantee of interchangeability.
| Project | Documented purpose | Platform | Access and requirements |
|---|---|---|---|
| Eve | Run iterative experiments that improve coding-agent solutions and guidance. | CLI | The repository identifies Eve as open source under the Apache-2.0 license. Running agents requires separately configured provider credentials; Codex is the default backend. |
| Dexto | Define agents in YAML with a model, tools, storage and permissions. | CLI |
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 7 October 2026
- Original website title
- scaling-group/eve · Official homepage title · retained in its source language