programmatic-tool-calling

General-purpose AI agent that acts via code actions through a unified execution interface.

Unified execution environment for Python code, shell commands, and programmatic MCP tool calls.

An Open Source Typescript implementation of Programmatic Tool Calling for AI Agents (Based on Code Mode)

⚡ Cut LLM inference costs 80% with Programmatic Tool Calling. Instead of N tool call round-trips, generate JavaScript to orchestrate tools in Vercel Sandbox. Supports Anthropic, OpenAI, 100+ models via AI Gateway. Novel MCP Bridge for external service integration.

Benchmarked agent execution runtime for Python. Sub-10ms cold starts, real-time streaming, time-travel debugging, and self-growing tool libraries. Compare 3 sandbox backends: Docker (OpenSandbox), MicroVM, and in-process AST.

Programmatic tool calling / Code Mode for MCP — turn any OpenAPI spec into two sandboxed tools (search + execute).

Programmatic tool calling orchestration for Pi powered by TypeScript

An MCP server implementation for Programmatic Tool Calling (PTC) that runs JavaScript

The first trainable, 100% domain-agnostic PyTorch Cortical MCP & Multi-Agent Engine. Wakes n (default=4) orthogonal specialists from a 1-sentence prompt, gates working memory at 0 routing tokens (-60% context), verifies designs in Python/SciPy before answering, and trains its own weights on the fly via Dream-RSI.

A notebook-first benchmark comparing Direct and Programmatic Tool Calling across deterministic inventory, incident investigation, and refund approval workflows, with quality gates, token usage, latency, trace analysis, and estimated cost.

Generate typed Python APIs from MCP server tool schemas

Programmable tool calling (PTC) and subagent fan-out for pi — the model writes JS/TS programs that call pi's tools, or fans out to a fresh pi subprocess per subagent task

Python SDK and LangChain / Deep Agents middleware for Submilli: code mode for agents, with TypeScript programs that run on submilli-server under a Blueprint's policy.

🤖 Optimize LLM usage by streamlining tool calls with code generation and sandbox execution for reduced costs and latency in any model.

DeepSeek Harness 高效执行 Skill：将确定性操作合并为单次脚本化执行、反馈压缩为摘要并按需深读、本地核验结果、及时停止无效重试 | Efficient tool execution skill for DeepSeek Harness: batches deterministic operations into one scripted pass, compresses feedback to summaries with on-demand deep reads, verifies results locally, and stops unproductive retry loops