altk-evolve

Name: altk-evolve Review
Item: altk-evolve
Rating: 3.2
Author: Aaron Matthews

Skill Development Usable

Works inClaude Code

Usable Scanned — metadata only

Reduces manual iteration cycles by enabling agents to refine outputs based on execution feedback. Cuts development time for complex task automation.

A system from IBM Research/AgentToolkit that gives AI coding agents memory so they learn from past sessions instead of repeating the same mistakes. It combines an MCP server, vector storage, and LLM-based conflict resolution to refine a knowledge base from agent trajectories. The Lite version slots into Claude Code and Codex; on the AppWorld benchmark it improved agent reliability by 8.9 points overall.

Agent refines API call logic after detecting malformed requests
Code generation agent improves output based on test failures
Data extraction agent adjusts parsing strategy when format validation fails

99 starsApache-2.0 (commercial OK)FreeQuick setup

Usable rating — This tool is functional but has notable gaps. Review the evaluation notes below before deploying.

Why it matters

Reduces manual iteration cycles by enabling agents to refine outputs based on execution feedback. Cuts development time for complex task automation.

Best for

Teams building multi-step AI workflows that need agents to self-correct without human intervention between steps.

Works in

Claude Code Claude Cowork Claude Chat

Source

https://github.com/AgentToolkit/altk-evolve

By AgentToolkit

How to Get It

Option 1: Claude Desktop App (Code Mode)Click the + button next to the prompt box → Plugins → Add plugin. Search and click Install. Skills work in Claude Code only.

Option 2: Paste into Claude CodeCopy the command below and paste it into your conversation. Claude will install it.

Command

pip install altk-evolve   (then run the MCP server with: uv run evolve-mcp)

Tip: Paste this into a Claude Code conversation. Verify command matches your Claude Code version.

First thing to try

After installing, paste this into Claude:

Help me agent refines API call logic after detecting malformed requests

PrerequisitesPython 3.12 or higher and uv (recommended) or pip. Optional semantic-search backends (pgvector or Milvus) and an OpenAI or LiteLLM key for LLM-based conflict resolution.CostFree

Trust Signals Auto-scanned

Stars99Contributors9Last updated2026-08-12LicenseApache-2.0 (OK for commercial use)Known CVEsNone foundSources: GitHub Advisory Database + OSV.dev · Scanned 2026-08-17 · scanner v1

Community Pulse Active

Discussed on Hacker News, Reddit

A long-term memory system that turns agent trajectories into reusable guidelines — Hacker News · 4 pts
ALTK‑Evolve: On‑the‑Job Learning for AI Agents — Hacker News · 2 pts
ALTK‑Evolve: On‑the‑Job Learning for AI Agents — Hacker News · 1 pts

3 mentions across 1 sources

Reviewer notes

Auto-scanned review. These are observations, not a security certification.

catalog_hygiene stale-eval refresh: Scored from trust signals (evidence-eval-v1): 99 GitHub stars; 9 contributors; last commit 6d ago; license Apache-2.0.

Things to check

Scanned, not hands-on tested — this entry was auto-scanned from public metadata (GitHub metrics, license, security flags). No reviewer has run it, and no tool-specific limitations have been documented yet.

How to evaluate tools before deploying →

Data shown here comes from public APIs and automated scanning. Reviewer notes reflect one person's experience. This is not a security certification or legal recommendation. Always evaluate tools according to your own organization's policies.

Evaluation Scores

Evaluation

Ease of Use

3/5

Versatility

3/5

Reliability

4/5

Security

3/5

Overall score3.25 / 5.00 UsableEvaluatedAug 2026

catalog_hygiene stale-eval refresh: Scored from trust signals (evidence-eval-v1): 99 GitHub stars; 9 contributors; last commit 6d ago; license Apache-2.0.

← Back to the Claude Observatory