# LLM Agents & Tool Use: State of the Field (October 2026)

Research dossier · v1 · updated 2026-10-30 · 3,336 words · 49 sources · 28 claims · verification: verified

Covers LLM agents that plan and call tools: function calling, coding and computer-use agents, multi-agent systems, benchmarks, security and runtime safeguards, as of October 2026. The key takeaway is that headline agent scores are fragile: benchmark leakage and weak tests can change what a reported number means, and interface drift can erode reliability after deployment, so evaluation design matters as much as model choice.

## Contents

- Executive summary
- Background and core concepts
- Timeline of key milestones
- State of the art
  - Tool calling and API use
  - Coding agents and software engineering
  - Computer-use and web agents
  - Multi-agent systems
  - Runtime safeguards and monitoring
  - Quantitative benchmarks
- Open problems and active debates
- Major programs and players
- Outlook: what to watch (next 12-18 months)
- Key papers
- Glossary

## Sample findings

- ReAct reports absolute success-rate gains of 34 and 10 points over imitation and reinforcement learning baselines, respectively, on ALFWorld and WebShop.
- Toolformer learns when and how to call external tools from self-generated supervision.

## Get the full dossier

GET https://space-kitty.com/api/v1/dossiers/ai-agents costs $0.20 via x402 (USDC on Base Sepolia (testnet)). Send `Accept: text/markdown` for Markdown or `application/json` for structured data (claims with citations, timeline, benchmarks, open problems, glossary, bibliography).
---
Space Kitty Archive · https://space-kitty.com · API: https://space-kitty.com/api/v1 · MCP: https://space-kitty.com/mcp · Premium: $0.20/request via x402 (Base Sepolia (testnet)). Metadata: arXiv (CC0), OpenAlex (CC0), Europe PMC (CC BY/CC0 works), ClinicalTrials.gov.