Route tasks to Google's Gemini models via local CLI agent. Uses your Google One AI Pro subscription as a fallback when Claude limits are reached, or for tasks benefiting from Gemini's capabilities.
MIT
204,280 tools. Last updated 2026-06-14 23:34
"LLVM" matching MCP tools:
- Override the last routing decision to correct wrong model choices and record feedback for routing quality improvement.MIT
- Scan workspace files (package.json, README, TODOs) to generate a compact semantic summary for routing context. Call at session start or after major refactors.MIT
- Routes AI tasks to the best model while tracking cumulative savings across sessions and hosts, even without client-side hooks.MIT
- Provide feedback on routing decisions to retrain the local classifier and improve model selection over time.MIT
- Displays real-time budget pressure across configured AI providers, using pressure bars to indicate quota exhaustion risk and help manage spending.MIT
Matching MCP Servers
- AlicenseAqualityAmaintenanceA local MCP server that packages LLM evaluation gates as reusable CI/CD primitives, enabling AI agents to run datasets against models, score responses, and enforce quality thresholds.Last updated10MIT
- AlicenseBqualityCmaintenanceEnables AI agents to interact with multiple LLM providers (OpenAI, Anthropic, Google, DeepSeek) through a standardized interface, making it easy to switch between models or use multiple models in the same application.Last updated16MIT
Matching MCP Connectors
Swiss weather data for AI assistants — forecasts, measurements, stations, pollen.
LLM provider intelligence: recommendations, pricing, and status
- Decide whether to proceed with a blocked high-cost AI call by approving, rejecting, or downgrading to a cheaper model.MIT
- Send a task to Codex CLI to run non-interactively with OpenAI models, bypassing Claude quota limits.MIT
- Classifies task complexity and routes to the optimal LLM. Uses a cheap classifier to select from budget, balanced, or premium models based on difficulty.MIT
- Evaluate and benchmark all available models on reasoning and code tasks to determine quality, speed, and accuracy, then optimize routing priorities.MIT
- Apply bulk edits across multiple files by generating JSON edit instructions from a natural language task. Use for cross-file refactors and pattern updates.MIT
- Open a local web dashboard to view real-time routing statistics, cost trends, model distribution, and recent decisions from the LLM Router. Refreshes automatically every 30 seconds.MIT
- Assess prompt complexity and token usage to select the best model, adjusting for budget pressure and quality needs.MIT
- Save a compact summary of the current session to provide context for future routed model calls, enabling cross-session awareness.MIT
- Converts natural language file search descriptions into glob/grep commands and executes them to locate files.MIT
- Routes your question to the most suitable LLM based on complexity or explicit model choice, balancing cost and capability.MIT
- Routes task prompts to the recommended agent and model by classifying complexity and applying your budget profile, enabling correct session-level setup.MIT
- Generate creative or long-form content by automatically routing your prompt to the optimal AI model based on task complexity. Supports writing, summarization, and brainstorming.MIT
- Automatically decompose complex tasks into multi-step pipelines across multiple LLMs, routing each step to the optimal model. Supports templates for common patterns or auto-decomposition.MIT
- Route coding tasks to the optimal AI model based on complexity for code generation, refactoring, and algorithm design.MIT