KodeAgent
VerifiedA minimal, framework-free engine for building reasoning AI agents.
What is KodeAgent?
KodeAgent is a compact Python library that serves as the reasoning engine inside larger applications rather than a complete agent platform. It implements ReAct, CodeAct, and native function calling while adding an internal planner and observer to keep tasks on track and recover from loops.
The library relies on LiteLLM for model access, allowing swaps between providers such as Gemini, GPT, Claude, or local Ollama instances. With few dependencies it runs in serverless settings, standalone scripts, or embedded within other codebases and is especially tuned for efficient small language models.
Developers who need a focused reasoning component without adopting a full framework will find it suitable. It is not intended for users seeking built-in persistent memory or turnkey task-specific agents.
What you can build with KodeAgent
Web research tasks
Combine search and page-reading tools with a ReAct agent to gather and compare information across sites.
Code-assisted problem solving
Use a CodeAct agent to generate and run Python snippets that handle calculations or data transformations on the fly.
SLM-powered agents
Deploy the FunctionCallingAgent with smaller models that have native tool-calling support for resource-constrained environments.
Install KodeAgent
pip install -U kodeagentpip install -U kodeagent # Upgrade existing installation- 1Install the package with pip install -U kodeagent.
- 2Import ReActAgent or CodeActAgent along with any desired tools.
- 3Instantiate the agent with a model name, tool list, and iteration limit.
- 4Call the async run method with a task description and stream responses.
- 5Access the final result via agent.task.result after the loop completes.
KodeAgent: pros & cons
Pros
- +Very small dependency footprint and easy to embed
- +Built-in planner and observer for plan correction
- +Supports ReAct, CodeAct, and native function calling
- +Works with many LLMs through LiteLLM without code changes
Cons
- –Still under active development so APIs may shift
- –No automated long-term memory persistence layer
- –Lacks the breadth of larger agent frameworks
Frequently asked questions
It supports ReAct for step-by-step reasoning, CodeAct for generating and executing code, and direct function calling.
User reviews
Verified reviews from the community shape this listing's rating.
Loading reviews…
Similar agents
Other general-purpose options worth comparing.
