ask-another
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@ask-another Ask GPT-4o about recent AI breakthroughs."
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
ask-another
An MCP server that gives your AI assistant access to other LLMs. Query hundreds of models across OpenAI, Google, and OpenRouter through a single interface — with model discovery, usage-based favourites, and automatic enrichment (Elo ratings, pricing, knowledge cutoffs). Includes deep-research jobs and image generation across multiple model families.
Install for Claude Desktop (CDA)
Install uv —
brew install uv(macOS / Linux with Homebrew) orwinget install astral-sh.uv(Windows). See docs/install.md for other platforms.Download the latest
ask-another.mcpbfrom Releases.Double-click the
.mcpbfile — Claude Desktop opens an install dialog and prompts for your API keys.Paste at least one provider key — OpenAI (
sk-…), Google AI Studio (Gemini), or OpenRouter (sk-or-…). Leave the others blank.Click Enable. First launch downloads Python dependencies via uv (~30s); after that, sub-second.
Related MCP server: OpenRouter MCP Server
Install for Claude Code (CC)
Install uv —
brew install uv(macOS / Linux with Homebrew) orwinget install astral-sh.uv(Windows). See docs/install.md for other platforms.Add the marketplace and install:
/plugin marketplace add matthewgjohnson/ask-another /plugin install ask-another@ask-anotherEnter your keys: open
/plugin→ ask-another → Configure options, paste at least one provider key, and Save. Run/reload-pluginsto apply.Optional: enable auto-update at
/plugin→ Marketplaces →matthewgjohnson/ask-anotherso future versions install on CC startup.
Other MCP clients (Cursor, Windsurf, …)
For clients without MCPB or plugin support, add the server manually. Example config:
{
"mcpServers": {
"ask-another": {
"command": "uvx",
"args": ["--from", "git+https://github.com/matthewgjohnson/ask-another", "ask-another"],
"env": {
"PROVIDER_OPENAI": "sk-your-openai-key",
"PROVIDER_GEMINI": "your-google-key",
"PROVIDER_OPENROUTER": "sk-or-your-openrouter-key"
}
}
}
}What You Can Do
"Ask GPT-5.2 what it thinks about this architecture" →
completion"What Gemini models are available?" →
search_models"Research the state of WebAssembly in 2026" →
start_research"Generate a logo for my project" →
generate_image"Note that DeepSeek is good for creative writing" →
annotate_models
See all 10 tools in the Reference.
Learn More
Installation Guide — Linux/Windows setup, Claude Code, Cursor, Windsurf, and other MCP clients, all configuration options
Reference — tools, annotations & enrichment, architecture, development
Building from source
git clone https://github.com/matthewgjohnson/ask-another
cd ask-another
make mcpb # builds ask-another.mcpbThen double-click the .mcpb to install in Claude Desktop.
License
Available Tools
10 toolsannotate_modelsA
Add or update a personal note on a model. Notes appear in search_models results and in the favourites list in server instructions.
Args: model: Full model identifier (e.g. 'openai/gpt-5.2'). note: Your note about this model. Overwrites any existing note.
Inputs: note (required), model (required). Returns: result
cancel_researchA
Cancel a running research task. Use check_research first to find the job_id of the task you want to cancel.
Args: job_id: The job to cancel.
Inputs: job_id (required). Returns: result
check_researchA
Check on research tasks started with start_research.
Called with no arguments, returns a table of all research tasks with their job_id, model, status, query, and timing. Called with a job_id, returns the full results of a completed task including the research report and cited sources.
Use this after start_research was interrupted or timed out, or to poll a long-running task.
Args: job_id: A specific job to retrieve. Omit to list all jobs.
Inputs: job_id. Returns: result
completionA
Call a model for a quick completion. Use this for standard prompts that return in seconds — use start_research instead for deep research tasks that need web search and source synthesis.
Use a favourite shorthand (e.g. 'openai') or an exact model ID verified via search_models. Shorthands and favourite model IDs listed in the server instructions can be used directly without calling search_models first.
Args: model: Full model identifier (e.g. 'openai/gpt-5.2') or favourite shorthand (e.g. 'openai' → resolves to your most-used OpenAI model). Use search_models to find other valid identifiers. prompt: The user prompt to send to the model system: Optional system prompt temperature: Sampling temperature (0.0-2.0). Omit to use model default. Some models reject non-default values — omit unless needed.
Inputs: model (required), prompt (required), system, temperature. Returns: result
feedbackARead-onlyIdempotent
Help us improve ask-another by sharing your experience. Call this whenever you're unsure how to proceed, receive confusing output, or a tool doesn't behave as expected. We also welcome suggestions — if a workflow felt more complex than it should be, if you had to guess at parameter values, or if something could simply work better.
Every piece of feedback helps us make ask-another more useful. This tool is lightweight and safe to call at any time.
Args: issue: Share what happened and what you expected. For suggestions, describe what could work better and why. tool_name: Which tool was involved, if applicable
Inputs: issue (required), tool_name. Returns: result
generate_imageA
Generate an image from a text prompt. The image is saved to disk (~/Pictures/ask-another by default), opened in the system default image viewer, and a preview is also returned in the tool result.
Where to view the generated image:
It pops open in your system image viewer automatically. Disable with OPEN_GENERATED_IMAGES=false if you don't want that.
The full-resolution file is at the path shown in the "Saved to:" message (open it directly for full quality).
In Claude Desktop, the inline preview lives inside the collapsed tool-use details — click to expand the tool call to see it. The chat does not render the preview as a top-level message.
Two model types are supported — the tool picks the right path automatically:
Dedicated image models (gpt-image-1, dall-e-3, imagen-4): best control over size and quality.
Native image-output models (gemini-*-image / "Nano Banana"): can interleave text and images, good for diagrams or annotated visuals.
Args: model: Model to use (e.g. 'openai/gpt-image-1', 'gemini/gemini-2.5-flash-image'). Use search_models with 'image' to find available image models. prompt: Text description of the image to generate. size: Image dimensions. Only used by dedicated image models — ignored by native image-output models. Common values: '1024x1024' (square), '1536x1024' (landscape), '1024x1536' (portrait). Valid options depend on the model. Omit to use the model's default. quality: Image quality. Only used by dedicated image models — ignored by native image-output models. For gpt-image-1: 'low', 'medium', 'high'. For dall-e-3: 'standard', 'hd'. Omit for the model's default.
Inputs: model (required), prompt (required), size, quality.
refresh_modelsA
Force a re-scan of all configured providers and re-fetch enrichment data from LMArena arena-catalog and LMArena metadata. Use this if model data seems stale or after adding a new provider.
Inputs: no named parameters. Returns: result
search_familiesA
Browse available provider groupings (e.g. 'openai', 'openrouter/deepseek'). Use this to explore what's available before drilling into specific models with search_models.
Args: search: Substring filter applied to family names zdr: Filter OpenRouter models to Zero Data Retention (ZDR) compatible only. Defaults to the server's ZERO_DATA_RETENTION setting. Set explicitly to override.
Inputs: zdr, search. Returns: result
search_modelsA
Find exact model identifiers. Always call this to verify a model ID before passing it to completion or start_research — do not guess IDs.
Args: search: Substring filter applied to full model identifiers zdr: Filter OpenRouter models to Zero Data Retention (ZDR) compatible only. Defaults to the server's ZERO_DATA_RETENTION setting. Set explicitly to override.
Inputs: zdr, search. Returns: result
start_researchA
Start a deep research task. This submits a research query to a model that will search the web, read sources, and synthesize a cited report.
Research tasks can take minutes to complete. This tool will wait for results and return them directly if the task finishes in time. If it is interrupted or the task times out, the research continues in the background — use check_research to retrieve results later.
Args: model: Model to use (e.g. 'openrouter/perplexity/sonar-deep-research'). Use search_models with 'deep-research' to find available models. query: The research question or topic to investigate. timeout: Max seconds to wait for results (default 300). If exceeded, the task continues in the background.
Inputs: model (required), query (required), timeout. Returns: result
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
TDQS
Scored across 10 tools
Each tool has a clearly distinct purpose: annotate_models manages notes, research tools handle lifecycle, completion for quick completions, generate_image for images, search tools for browsing providers/models, and feedback for user input. No overlapping functions.
Most tools follow a verb_noun pattern (e.g., search_models, start_research), with exceptions like 'completion' and 'feedback' being single words. The naming is mostly consistent and readable, using lowercase and underscores for compounds.
10 tools is well within the ideal range for a focused server. Each tool serves a distinct function without redundancy or clutter, covering model interaction, research, image generation, and feedback.
The tool set covers the full lifecycle for research tasks (create, check, cancel), model discovery (search_families, search_models), annotations, image generation, and user feedback. No obvious gaps are present for the server's stated purpose.
Maintenance
Related MCP Connectors
Image, video, music and text generation across 100+ models through one endpoint.
AI agent tools: web search, browser, 400+ LLMs, image gen, TTS, phone verify. Pay-per-use.
run any ai model. compose agents, stack knowledge, connect tools. one api, pay per run.
Image and video AI tools and your own pipelines, run from any AI assistant.
Related MCP Servers
- AlicenseAqualityFmaintenanceProvides AI assistants with real-time access to 1000+ AI models including their latest pricing, context windows, capabilities, and specifications. Supports model search, comparison, recommendations, and live testing.31MIT
- AlicenseNot gradedqualityDmaintenanceProvides seamless access to 200+ AI models through OpenRouter's unified API, featuring multi-model collaboration, vision support, intelligent benchmarking, and collective intelligence capabilities for enhanced decision-making.1,013 npm10MIT
- AlicenseAqualityBmaintenanceProvides access to 400+ AI models from OpenRouter, enabling users to chat with models like GPT-4, Claude, Gemini, and Llama, compare responses across multiple models, and retrieve model information with pricing details.489 npm14MIT

Adorbis AI MCP Serverofficial
AlicenseNot gradedqualityDmaintenanceEnables AI assistants to access 40+ models via one API key with automatic routing for chat, code, reasoning, and writing tasks.51 npm1MIT