[Feature]: Programmatic Tool Calling (PTC) for code based tool execution

[kemier]

When using tool calling with LLMs, each tool invocation requires a separate model turn, leading to:

High latency for multi-tool workflows
Increased token consumption from repeated context
Limited ability to perform complex logic (filtering, aggregation, conditionals) across tool results
xref:

Anthropic's https://www.anthropic.com/engineering/advanced-tool-use
Anthropic's https://www.anthropic.com/engineering/code-execution-with-mcp
MCP Server with Code Execution https://github.com/harche/ProDisco
Proposed Feature: Programmatic Tool Calling (PTC) allows models to write Python code that orchestrates multiple tool calls within a secure sandbox. Instead of returning individual tool calls, the model generates code that:

Calls multiple tools programmatically
Processes results with Python logic (loops, filters, aggregations)
Returns only the final result to the model context

flow is like following:

LLM writes in sandbox:
result = await search_files(pattern="*.py")

↓ (calls)
Generated wrapper function (tools/filesystem.py):
async def search_files(pattern: str) -> dict:
return await mcp_client.call_tool(
server_id="filesystem-server-001",
tool_name="searchFiles",
arguments={"pattern": pattern}
)

↓ (JSON-RPC 2.0 over stdio/HTTP)
MCP Server Process
↓ (returns result)
Wrapper deserializes response
↓
LLM receives Python dict

this could help to save a lot of tokens.