Skip to content

Kodiqa Agent

v3.8.0 Feature

This release adds 3 notable features for engineering teams evaluating rollout.

Published 1mo AI Agents & Assistants
✓ No known CVEs patched
Read the diff → Tool health → What is this tool? →

✓ No known CVEs patched in this version

Topics

ai-coding-agent claude cli coding-assistant developer-tools ollama
+2 more
openai python

Summary

AI summary

Lazy MCP tool loading reduces per-turn token cost by ~94% for large servers.

Changes in this release

Feature Medium

Adds lazy MCP tool loading (on by default).

Adds lazy MCP tool loading (on by default).

Source: llm_adapter@2026-06-07

Confidence: high

 —
Feature Medium

Introduces three fixed meta-tools: mcp_search, mcp_tool_schema, and mcp_call.

Introduces three fixed meta-tools: mcp_search, mcp_tool_schema, and mcp_call.

Source: llm_adapter@2026-06-07

Confidence: high

 —
Feature Medium

Implements usage‑aware ranking in mcp_search based on call frequency stored in ~/.kodiqa/mcp_usage.json.

Implements usage‑aware ranking in mcp_search based on call frequency stored in ~/.kodiqa/mcp_usage.json.

Source: llm_adapter@2026-06-07

Confidence: high

 —
Feature Medium

Adds `/mcp lazy [on|off]` command to toggle lazy loading and shows token trade‑off in `/mcp list`.

Adds `/mcp lazy [on|off]` command to toggle lazy loading and shows token trade‑off in `/mcp list`.

Source: llm_adapter@2026-06-07

Confidence: high

 —
Feature Medium

Allows disabling lazy loading via `mcp_lazy: false` setting to retain always‑inject behavior.

Allows disabling lazy loading via `mcp_lazy: false` setting to retain always‑inject behavior.

Source: llm_adapter@2026-06-07

Confidence: high

 —
Feature Low

Adds `test_mcp_lazy.py` with 383 tests covering lazy vs. non‑lazy exposure, meta‑tool handlers, usage ranking, and routing.

Adds `test_mcp_lazy.py` with 383 tests covering lazy vs. non‑lazy exposure, meta‑tool handlers, usage ranking, and routing.

Source: llm_adapter@2026-06-07

Confidence: high

 —
Performance Medium

Reduces per‑turn tool‑schema token cost from ~5,300 to ~310 tokens (~94% reduction) for a 50‑tool server.

Reduces per‑turn tool‑schema token cost from ~5,300 to ~310 tokens (~94% reduction) for a 50‑tool server.

Source: llm_adapter@2026-06-07

Confidence: high

 —
Full changelog

Lazy MCP tools — large MCP servers no longer cost you tokens every turn. Inspired by mcp2cli and Anthropic's Tool Search.

Added

  • Lazy MCP tool loading (on by default). Instead of injecting every connected MCP tool's schema into every request, Kodiqa exposes 3 fixed meta-tools and discovers tools on demand:

    • mcp_search — find MCP tools by keyword (names + descriptions, ranked by how often you've used them)
    • mcp_tool_schema — fetch one tool's full input schema only when needed
    • mcp_call — execute a tool by name

    For a 50-tool MCP server this cuts the per-turn tool-schema cost from ~5,300 to ~310 tokens (~94%), and the cost stays flat no matter how many MCP tools are connected.

  • Usage-aware ranking — mcp_search orders results by call frequency (persisted in ~/.kodiqa/mcp_usage.json).

  • /mcp lazy [on|off] — toggle it; /mcp list shows the current mode and token trade-off. Set mcp_lazy: false in settings to keep the old always-inject behavior.

Tests

  • test_mcp_lazy.py — manager primitives, lazy vs. non-lazy tool exposure, meta-tool handlers, usage ranking, and routing. 383 total; ruff + CI green.
View diff on GitHub

Weekly OSS security release digest.

The CVE patches and breaking changes that affected production tools this week. One email, every Sunday.

No spam, unsubscribe anytime.

Share this release

Share on X Share on Bluesky

Track Kodiqa Agent

Get notified when new releases ship.

Sign up free

About Kodiqa Agent

All releases →

Related context

Beta — feedback welcome: [email protected]