Glossary · Term

tool call

← all terms

Definition

Plain language

A structured request an AI agent makes to use an external function or service.

As stated in the literature

A structured invocation issued by an LLM agent to call an external function, tool, or API, whose result is fed back into the model's context.

Also called: tool calls, tool use

Why it matters: Tool calls are how LLMs reach beyond their training data to take real actions, and their correctness is central to every agent's reliability.

For example, an agent answering 'what's the weather in Tokyo' issues a tool call to a weather API with city='Tokyo' and uses the returned JSON to compose its reply.

Heard on the show

“In a real deployed system, actions are tool calls — actual code, actual transactions — with observable consequences the monitor can check against ground truth.”
Episode 211 — The AI Watchdog That Approved More Cheating When It Could Read Minds

Mentioned in 51 episodes

  1. 211
    The AI Watchdog That Approved More Cheating When It Could Read Minds
  2. 189
    Why Phone Agents Ace the Test and Crash on Your Actual Phone
  3. 187
    An 8-Billion Agent That Beats Models 80 Times Its Size By Looking Things Up
  4. 184
    An AI Built an Undetectable Secret Channel, And Another AI Couldn't Find It
  5. 183
    Why You Can't Fine-Tune Foresight Into an AI Agent
  6. 181
    How to Backpropagate Blame Through a Team of Chatbots — And When It Backfires
  7. 175
    One Crosscoder Feature Flips a Stalling Chatbot Into a Working Agent
  8. 168
    When Turning Experience Into Code Makes Your AI Agent Dumber
  9. 167
    How Teaching an AI to Predict, Not Act, Made It a Better Actor
  10. 164
    The Summarizer That Quietly Deletes Your Agent's Safety Rules
  11. 162
    The Empty-Lake Proof: Why More Rollouts Stop Helping Reasoning Models
  12. 146
    How an Innocent README Can Freeze an AI Agent's Safety Check for an Hour
  13. 144
    When an AI Agent Just Copies Its Tool — And Bigger Models Copy More
  14. 126
    How Coding Agents Can Mine Their Own Failures Into a Self-Targeting Curriculum
  15. 125
    AI Coding Agents Run a Marathon, and Fewer Than One in Three Finish
  16. 123
    Five Identical Worlds, One Swapped Model: What Happens When AI Agents Run for Fifteen Days
  17. 121
    When the Agent Says It's Done But Nothing Happened: Debugging the Harness, Not the Model
  18. 120
    How an AI Agent Rewrites Its Own Tools, Without an Answer Key
  19. 109
    An AI Got Caught Reading the Answer Key, And Why That Catch Matters
  20. 105
    The Trojan Is Your Agent's Memory: Why Single-Step Defenses Miss Persistent Attacks
  21. 097
    Same Tokens, Same Cost, Wildly Different Results: What Actually Scales in AI Agents
  22. 096
    How Treating an AI Agent's Execution Like Git Recovers a Coordination Penalty
  23. 094
    Chain-of-Thought Monitoring Fails Across Languages, and Worst Where It's Needed Most
  24. 090
    How MiniMax-M2 Bets That Sparsity Plus Verifiable Rewards Can Match Frontier Agents
  25. 088
    Two Levers for Self-Improving AI: When Rewriting Code Isn't Enough
  26. 082
    Training a Deep Research Agent on 8,000 Synthetic Tasks: The Rubric Tree Trick
  27. 080
    How a Two-Agent Trick Unlocked Large-Scale Training for Computer-Use Agents
  28. 076
    Same Model, Organized Differently: How an Agent Architecture Beat Frontier Systems at Research Math
  29. 072
    A Robot Made Graphene Without Help, And Caught Itself Hallucinating
  30. 071
    When the Model Is Fine and the Plumbing Is Broken: Fixing Agents at the Interface
  31. 066
    Why Giving an AI Agent More Tools Can Make It Worse at Using a Computer
  32. 063
    Why Web Agents Are Slow: A Compiler-Style Fix for Computer-Use Latency
  33. 059
    Firefly's Inversion: Building Verified Tool-Call Training Data by Working Backward
  34. 057
    How Uber Caught 206 Leaked Credentials With an LLM-Powered Security Stack
  35. 047
    When Agent Benchmarks Lie: The Harness Problem in Open-Source AI
  36. 046
    When the AI Optimizer Edits the Grade Book: Why Harnessing Evolution Needs a Wall
  37. 044
    How One Sentence and a Forged History Flip the Most Aligned Models
  38. 040
    Two Frozen Models Learn to Whisper: Coupling Through Hidden States
  39. 039
    When Smarter Agents Get Fooled by Three Extra Nodes in a Database
  40. 035
    Why Frontier Agents Ask for Clarification at Exactly the Wrong Moment
  41. 034
    Catching Multi-Agent Deadlocks Before Deployment With a 40-Year-Old Tool
  42. 029
    Why Forty-Eight Percent on FrontierMath Isn't the Real Story in DeepMind's New Math Paper
  43. 021
    Ten Thousand Examples Beat the Full Industrial Pipeline for Search Agents
  44. 020
    The Compliance Gap: Why AI Says Yes and Does No
  45. 016
    Why Your Coding Agent Stalls While the GPU Runs Hot
  46. 014
    Why a Constrained Pipeline Beat a Full Coding Agent at Finding Bugs 30-to-1
  47. 013
    Why Search Keeps Rediscovering the Same Workflow, and What That Means
  48. 011
    When RL Actually Teaches Agents Something New, And When It Doesn't
  49. 008
    Why Long-Horizon AI Agents Get Stuck, and a Milestone-Based Fix That Helps
  50. 002
    An AI Ran a Real Optics Lab for 21 Hours and Found a Transformer-Shaped Pattern in Light
  51. 001
    When AI Models Quietly Protect Each Other From Shutdown

Related concepts

BrowseComp LLM Agents Structured Trace Formatting

Related terms

agent API