Open Source

Teotl: our open-source autonomous agent framework

Teotl is a production-focused framework for building AI agents that can run real, long-running tasks on their own — and stay safe doing it. Free to use, inspect, and extend under the MIT license.

Smart planning, low-cost execution

Teotl's planner-worker design pairs a strategic model (Claude Sonnet) that decides what to do with a fast, low-cost model (Claude Haiku) that does the work — cutting cost roughly 40% versus single-model approaches in our internal benchmarks, without giving up quality.

What makes it different is what most agent frameworks skip: security and control. Teotl ships with built-in guardrails that block dangerous commands, require human approval for sensitive actions, and track cost against a budget — plus credential management that keeps secrets in your OS keyring or AWS Secrets Manager, never in code. It's the human-at-the-seams principle we bring to every engagement, written into the software itself.

What's inside

Planner-worker architecture

A strategic model plans, a fast model executes — about 40% lower cost in our internal benchmarks.

Built-in guardrails

Blocks risky commands, requires approval for sensitive operations, and enforces budgets.

Secure by design

Credentials live in your OS keyring, encrypted storage, or AWS Secrets Manager — never in code.

Autonomous missions

Long-running workflows that execute without supervision, with built-in error recovery.

Modular skills

Filesystem, git, bash, and Python out of the box — plus your own custom skills.

Open source (MIT)

Free to use and extend. Python 3.11+. Works with Claude or GPT on macOS, Linux, and Windows.

What teams use it for

  • DevOps automation — autonomous bug investigation, test fixing, deployment monitoring
  • Code review — automated PR reviews and best-practice enforcement
  • Research & analysis — multi-step research and data gathering
  • Testing — generating test cases and finding edge cases
  • ETL & data pipelines — extraction and transformation with error recovery

Get started

Install from source (PyPI package coming soon):

git clone https://github.com/keithdit4e/teotl
cd teotl
pip install -e .

Full documentation, examples, and a 5-minute quick start are in the GitHub repository.

Get announcements

Subscribe with your email to hear when we publish open source projects, releases, and updates.