wende/cicada

Overview

A lightweight Python tool designed to provide guidance cues for AI coding agents that search without sufficient context. It helps structure or reroute search queries to improve relevance and reduce blind exploration.

Best for

Best for
Developers building or extending AI coding agents who need a simple way to inject search direction.

Use cases

• Injecting contextual hints into AI code search pipelines
• Reducing hallucination in autonomous coding agents by narrowing search scope
• Adding guided search logic to existing agent frameworks

How to use

Install

uvx cicada-mcp claude   # or cursor, or vs

Tools exposed

• uvx
• cicada-mcp
• claude
• cursor
• vs
• gemini
• codex
• opencode
• zed
• kimi

Tested with

Claude Code, Cursor, VS Code, ChatGPT

Example client config

{\n  "mcpServers": {\n    "cicada": {\n      "command": "cicada-mcp",\n      "args": ["--watch"],\n      "env": {\n        "CICADA_CONFIG_DIR": "/home/user/.cicada/projects/<hash>"\n      }\n    }\n  }\n}

Notes

A lightweight Python tool designed to provide guidance cues for AI coding agents that search without sufficient context. It helps structure or reroute search queries to improve relevance and reduce blind exploration.

37 stars on GitHub. Last updated 2026-03-03. Licensed MIT.

Use cases

• Injecting contextual hints into AI code search pipelines
• Reducing hallucination in autonomous coding agents by narrowing search scope
• Adding guided search logic to existing agent frameworks

Pros

• Lightweight and easy to integrate into existing Python agent codebases
• Focused on a specific pain point (blind search) without feature bloat
• Open source with a permissive license allows customization

Cons

• Very small community (37 stars) limits support and contributions
• Documentation and examples are minimal, requiring trial-and-error adoption
• Effectiveness depends heavily on how guidance cues are defined and tuned

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.