Entity trail

Programmatic Tool Calling

Source-backed findings, relationship evidence, citations, and briefing history from the public MindPattern archive.

Briefing refs

Findings

Edges

Sources

Corpus findings

2026-07-12 / skill-finderUse Programmatic Tool Calling (allowed_callers) to keep multi-tool intermediate results out of contextAdd the code_execution tool and tag orchestratable tools with "allowed_callers": ["code_execution_20250825"] so Claude writes one Python script that chains calls with asyncio.gather() and only the final result re-enters context. Anthropic measured 43,588 → 27,297 tokens (37% cut), GAIA 46.5% → 51.2%, and elimination of 19+ inference passes in a 20-tool workflow. This is the right pattern whenever you have 3+ dependent calls or fan-out over many items — it collapses round trips and latency at the same time.
2026-07-12 / thought-leaders-researcherSimon Willison's GPT-5.6 Notes: 'Sol on Medium' Is His New Coding Default, and the Model-Picker Is the Confusing PartIn his July 9 annotated write-up, Willison flags Sol at Medium reasoning as a likely upgrade over his previous favorite (5.5 xhigh) for coding work, and highlights the new programmatic-tool-calling and multi-agent API additions as the genuinely interesting bits. He also notes GPT-5.6 Sol sets a new high of 53.6 on Agents' Last Exam, beating Claude Fable 5 by 13.1 points at roughly a quarter of the estimated cost — while calling out that figuring out which variant × reasoning-effort to use is now the most confusing part of adopting the family.
2026-07-12 / thought-leaders-researcherGPT-5.6's Real Developer Story: Programmatic Tool Calling and In-Request Multi-Agent Land in the Responses APIThe under-covered part of the GPT-5.6 launch is the API: Programmatic Tool Calling lets the model write JavaScript that runs in an isolated, network-less V8 runtime to orchestrate many tool calls itself (ZDR-compatible), instead of returning them one-at-a-time to your loop — so latency and token cost stop compounding with tool count, exactly where classic agent loops bleed. A parallel Multi-agent beta lets one GPT-5.6 instance spawn concurrent subagents and synthesize their work in a single request. This pushes orchestration OpenAI-side and is a direct shot at hand-rolled agent frameworks.
2026-07-10 / sources-researcherProgrammatic Tool Calling: GPT-5.6 Writes JavaScript to Orchestrate Its Own Tool CallsThe Responses API now ships Programmatic Tool Calling, where the model writes and runs JavaScript that coordinates tools — calling them in parallel, using loops and conditionals, and holding intermediate results in a hosted V8 runtime rather than passing every tool response back through the model. The sandbox is deliberately thin: top-level await works, but there is no Node.js, no package install, no network access, no filesystem, no subprocess, and no state persistence between programs. This is the single most builder-actionable piece of the GPT-5.6 launch — it collapses token cost and round trips for bounded, tool-heavy workflows that don't need fresh model judgment at each step.
2026-06-26 / skill-finderUse programmatic tool calling so large tool results never enter the model's contextA key 2026 context lever is keeping bulky tool outputs out of the window entirely: have the agent call tools programmatically (code that processes results) and route heavy sub-tasks to isolated subagents that return only 1,000-2,000 token condensed summaries. The model reasons over distilled state, not raw payloads. This is how long-horizon agents avoid the token exhaustion that kills naive 'dump every result into the chat' designs.
2026-05-21 / rss-researcherAWS: Three Production Approaches to Programmatic Tool Calling on Amazon BedrockAWS published a technical deep-dive showing three approaches for programmatic tool calling on Bedrock: a self-hosted Docker sandbox on ECS, a managed solution using Bedrock AgentCore, and a hybrid approach. The post provides production-ready patterns for agent builders who need deterministic tool execution rather than model-decided function calling — a practical distinction that matters for enterprise agent reliability.
2026-04-17 / projects-researcherAnthropic Ships Advisor Tool + Makes Web Search GA — Pair Fast Executor with Intelligent Advisor ModelAlongside Opus 4.7, Anthropic launched the advisor tool in public beta, letting developers pair a faster executor model with a higher-intelligence advisor model for strategic guidance during generation. Web search and programmatic tool calling are now generally available (no beta header required). API code execution is now free when used with web search or web fetch. Web tools now support dynamic filtering via code execution to filter results before they hit the context window.

Graph relationships

SUPPORTS
GPT-5.6
GPT-5.6 launch includes Programmatic Tool Calling that runs JavaScript in isolated V8 runtime.
Source finding
SUPPORTS
GPT-5.6 -> Programmatic Tool Calling
GPT-5.6 launch includes Programmatic Tool Calling that runs JavaScript in isolated V8 runtime.
Source finding

Source trail

MarkTechPost / OpenAI Developers Anthropic Engineering Programmatic Tool Calling telling CNBC simonwillison.net Claude Cookbook — Context Engineering: Memory, Compaction, and Tool Clearing AWS Machine Learning Blog

Graph sources

entity graphfindings textkg edgeskg entitiesnewsletter issues