#149149th of 153 in Agent frameworks and SDKs
deepresearchagent
Python runtime for agents that version resources and self-improve
- Stars
- 3.6k
- License
- MIT
- Last commit
- May 2026
- Last release
- Feb 2026
- Language
- Python
3,311 on 25 Apr 20263,572 on 14 Nov 2026
From GitHub's daily star counts.
Checked on GitHub.
Overview
Deep Research Agent defines two protocol layers: RSPL registers prompts, agents, tools, environments and memory as versioned resources, and SEPL proposes, assesses and commits changes with lineage and rollback. It ships a tool-calling agent plus optimizers (reflection, GRPO, Reinforce++), tracing and MMEngine-style configs. The README documents only one example script, run with an OpenRouter-style model name.
Who it is for: Researchers and engineers building self-improving agent systems in Python
Strengths
- Versioned resources with rollback and auditable lineage for prompts, tools and memory
- Pluggable optimizers including reflection, GRPO and Reinforce++
- Structured traces and memory events help analyze failed runs
- MMEngine-style configs compose agents, tools, environments and models
Weaknesses
- No Docker or compose files provided
- README gives no benchmarks or results for the self-evolution loop
- Protocol terminology raises the learning curve for simple agent use
What it needs
- Needs LLM provider API key (e.g. OpenRouter)
- Models: openrouter/gpt-4o
What changed
From the release notes. No flag means the note does not say so, not that the upgrade is safe.
Also in Agent frameworks and SDKs
See all 153| Rank | Project | Score |
|---|---|---|
| 1 | ppt-masterAgent skill that turns documents into natively editable PowerPoint decks | 96 out of 100 |
| 2 | copilotkitReact and Angular SDK for adding agent chat and generative UI to apps | 92 out of 100 |
| 3 | openai-agents-pythonPython SDK for multi-agent workflows with handoffs, guardrails and tracing | 92 out of 100 |
| #4 | mastraTypeScript framework for building AI agents, workflows and MCP servers | 92 out of 100 |
| #5 | browser-usePython library that lets an LLM agent control a web browser | 90 out of 100 |