universal-ai-agent-bridge
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
Instructions
Guidance the server publishes about itself, which clients place ahead of the tool catalog so the model reads it before choosing anything.
This server publishes no instructions, or was last inspected before Glama recorded them.
Capabilities
Features and capabilities supported by this server
Protocol revision2025-11-25
| Capability | Details |
|---|---|
| tools | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| llm_queryA | Query ANY external LLM using an OpenAI-compatible /v1 endpoint (e.g. Dahl, DeepSeek, OpenRouter, Groq, Ollama, LM Studio, FreeLLMAPI, 9Router, OmniRoute, or any custom endpoint). Supports reasoning/thinking extraction, custom system prompts, and ad-hoc endpoint overrides. |
| llm_list_modelsB | List available models from an OpenAI-compatible /v1 endpoint by calling GET /v1/models. |
| llm_test_connectionA | Test connectivity, latency, and credentials for any LLM endpoint or configured provider. |
| llm_manage_providersB | Manage configured LLM providers in your Antigravity vault. List all providers, add a new endpoint, set active default, or update API keys. |
| llm_compareB | Benchmark and compare responses to the exact same prompt across multiple models or providers in parallel (e.g. Dahl vs DeepSeek vs Groq vs Ollama). |
| llm_generate_imageA | Generate an image using ANY external /v1 endpoint that supports image generation (e.g., SiliconFlow, Together AI, OpenAI DALL-E, OpenRouter, or local ComfyUI/Diffusers bridge). Saves the generated image locally and returns an embeddable file path. |
| llm_generate_videoC | Generate a video using an external /v1 endpoint that supports video generation (e.g., SiliconFlow LTX-Video, Together, MiniMax Hailuo Video, or custom video bridge). |
| llm_autodetectA | Auto-scan local network ports for running AI servers (Ollama on 11434, LM Studio on 1234, 9Router/OmniRoute on 20128, FreeLLMAPI on 4000, vLLM on 8000, LocalAI on 8080, Jan on 1337). Detects online engines, lists their active models, and automatically syncs them into your Antigravity vault. |
| llm_compress_promptA | Compress any prompt, code block, log output, or stack trace using RTK Smart Compression to strip redundant whitespace, duplicate lines, and deep library frames while maintaining code semantics. Returns token savings metrics. |
| llm_councilA | Multi-Model Consensus & Council Deliberation. Broadcasts a complex problem, code architecture, or security audit in parallel to multiple LLM council members (e.g. Dahl, DeepSeek, Groq, Ollama), then automatically uses a designated Chief Justice / Synthesizer model to critique, cross-examine, and deliver the single optimal consensus solution. |
| llm_get_analyticsA | Retrieve comprehensive cost, token usage, and latency analytics from the Antigravity Local LLM Ledger. View total tokens spent, estimated USD expenditure, breakdowns per provider (Dahl, Groq, DeepSeek, Ollama) and per model, or reset ledger statistics. |
| llm_presetsB | Manage and inspect expert System Personas and Prompt Presets (e.g. 'security-auditor', 'systems-architect', 'code-simplifier', 'quant-trader', 'fullstack-reviewer', 'explain-like-pro'). List available personas, view system prompts, or save custom domain presets to the vault. |
| llm_cacheA | Manage the Dynamic Response Cache for external LLMs. Inspect cache hit rates, view token and USD savings, prune expired entries, or clear the cache. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
No resources | |
TDQS
Scored across 13 tools
Most tools have clear purposes, but llm_query, llm_list_models, and llm_test_connection could be confused - a user might call query when they only want to list models, or test_connection when checking availability. llm_generate_video and llm_generate_image are distinct by media type, but llm_compare, llm_council, and llm_query all involve sending prompts to multiple models, though their outputs differ significantly.
All tools follow a consistent 'llm_' prefix with descriptive verb_noun naming (generate_video, list_models, test_connection, manage_providers). Minor inconsistency: 'llm_autodetect' is a verb alone without an object, and 'llm_cache' is a noun-only name rather than a verb_noun pair, but overall the pattern is predictable and readable.
13 tools is well within the ideal range for a domain-focused server. The tools cover distinct operational areas: querying, generation, model management, caching, analytics, and multi-model orchestration, each earning its place without unnecessary duplication.
The surface covers core LLM operations: querying, listing models, testing connections, managing providers, generating content, comparing models, and caching. Missing operations like streaming responses, stopping/canceling ongoing requests, or batch processing are minor gaps for a local dev tool, but the main workflows are complete.