---
id: tom-swe
version: "1.0.3"
license: MIT
license_treatment: permissive
maintenance: abandoned
---

# tom-swe — Theory of Mind modeling for Software Engineering assistants

License: permissive · Maintenance: abandoned · Downloads: 9.2M/mo

## What it is and what it does

TOM-SWE is a Theory of Mind package that augments software engineering agents with user understanding capabilities. It analyzes user behavior, preferences, and working styles through LLM-powered consultation, building a three-tier memory system from session data to user profiles. The package integrates with OpenHands and similar SWE agent frameworks, allowing agents to provide personalized guidance and adapt their behavior based on learned user intent.

The package depends on jinja2, json-repair, litellm, pydantic, python-dotenv, tiktoken, and tqdm to handle templating, LLM communication, configuration, and progress tracking. It requires an external LLM API (configured via litellm credentials) to perform the psychological analysis and consultation generation that drives its core functionality.

Use it for:

- Enhance OpenHands or other SWE agents with automatic user preference learning and personalized task guidance.
- Analyze user session data to extract behavioral patterns and working style insights for agent adaptation.
- Provide consultation recommendations to agents based on learned user psychology and preferences.
- Build user profiles from cleaned session data to enable long-term personalization across multiple interactions.
- Integrate Theory of Mind reasoning into autonomous coding agents for improved task performance.

## Worth the install?

AI-flagged interpretation of the facts above — verify before relying.

Adds Theory of Mind capabilities to software engineering agents, enabling them to understand and adapt to individual user preferences, working styles, and intent through LLM-powered user modeling and consultation.

No. The repository is archived and marked abandoned, indicating no active maintenance or support. While the package itself has low install friction and a permissive MIT license, the lack of ongoing development creates risk for bug fixes, security updates, and compatibility with evolving dependencies. Use only if you are willing to maintain a fork or accept the possibility of unresolved issues.

## Install

pip install tom-swe
uv add tom-swe
poetry add tom-swe

## Installing tom-swe

Before you install: Low friction installation as a pure Python wheel. Maintenance is a concern: the repository is archived and marked abandoned as of the fact sheet date, with no active development despite recent release dates in the metadata.

License in practice: MIT license permits commercial and private use with minimal restrictions, making it freely usable in most contexts.

Quickstart:

pip install tom-swe

from tom_swe.tom_module import TomModule
import asyncio

async def demo():
    tom = TomModule()
    consultation = await tom.consult(
        user_id="demo_user",
        current_context="User wants to implement a new feature"
    )
    print(consultation)

asyncio.run(demo())

Requires Python 3.10 or later. Requires LLM API credentials (litellm key and endpoint) configured via .env file to function.

Verify before relying:

- Whether the archived repository status means the package is no longer maintained or supported.
- How the three-tier memory system (sessions → analyses → profiles) performs with real user data at scale.
- Compatibility details with specific OpenHands versions or other SWE agent frameworks beyond OpenHands.

## Package facts

- License: MIT (permissive)
- Python support: supports_current
- Install friction: low
- Maintenance: abandoned
- Downloads: 9.2M/month (top 5,000 on PyPI)
- Known vulnerabilities: none known

## Tags

user modeling for AI agents, theory of mind software engineering, personalized agent behavior, user preference learning agents, adaptive SWE agent framework, LLM user psychology analysis, agent consultation system, user-modeling, agent-framework, llm-integration

[View on SkillFed](https://skillfed.io/packages/tom-swe) · [View on PyPI](https://pypi.org/project/tom-swe/)