Kapso: Long-Running AI Agent Framework

Understand the project

Practical overview

When it fits

Use Kapso when the task benefits from repeated experimentation, measurable progress and accumulated experience rather than a one-off generation step.

Who it is for

Teams building long-running coding-agent systems.
Researchers prototyping AI or data-system optimization workflows.
Developers exploring program synthesis, knowledge graphs or benchmark-oriented machine-learning agents.

Documented uses

Run iterative experiments toward a stated software or AI-system objective.
Generate, implement and compare candidate solutions.
Retain evidence-backed lessons from completed campaigns for later work.
Prototype optimization workflows for machine-learning and data systems.

Before you choose

Requirements & limits

The repository is public and identifies an MIT license. Kapso is published on PyPI as leeroo-kapso, and project documentation is linked at https://docs.leeroo.com/docs.

Whether the project requires iterative campaigns rather than single-step agent output.
Whether retaining lessons and knowledge across campaigns is useful.
Whether the team can validate deployment requirements, since the available description does not specify the execution architecture.
Whether Python is an appropriate implementation environment.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install Kapso and check its prerequisites

Install the Python package and verify which configured requirements are present.

Before you start

Node.js 18 or later.
Python 3.10 or later.
The Codex and Claude Code CLIs installed globally and logged in, as listed in the Quickstart.

Steps

Install and log in to the Codex CLI using the command below and codex login.
Documented command
```
npm install -g @openai/codex
```
Install and log in to Claude Code using the command below and claude auth login.
Documented command
```
npm install -g @anthropic-ai/claude-code
```
Install Kapso with the command below. The package is named leeroo-kapso; the unrelated PyPI package named kapso can shadow its command.
Documented command
```
pip install leeroo-kapso
```
Run kapso doctor to check the setup, or kapso doctor evolve to check requirements for that verb.

Expected outcome

Kapso is installed, and its doctor command reports configured requirements and names fixes for missing items.

Guide limitations

Kapso uses coding-agent CLIs for inference and has no direct-API fallback.
An OpenAI key is optional; it is used for embeddings only when the knowledge graph or semantic search over past experiments is enabled.
Requirements depend on configuration, and some optional features may require Docker, Weaviate, Neo4j, or deployment targets.

Documentation 1 ↗Documentation 2 ↗

COMPARE OPTIONS

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
Kapso	Run iterative experiments toward a stated software or AI-system objective.	Cloud, Self-hosted, Server	The repository is public and identifies an MIT license. Kapso is published on PyPI as leeroo-kapso, and project documentation is linked at https://docs.leeroo.com/docs.
Dexto	Define agents in YAML with a model, tools, storage and permissions.	CLI

Evidence and transparency

Sources & review

Official repository ↗Guide documentation ↗

Source record details

Source availability: Active
Sources checked: 30 September 2026
Original website title: Leeroo-AI/kapso · Official homepage title · retained in its source language

How we review sources Suggest a correction

Kapso