codex-gemini-bridge
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| AGY_BIN | No | Path to the Antigravity CLI binary. If not set, the server will attempt to use 'agy' from PATH. |
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 | {
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| geminiA | Send a prompt to the local Antigravity CLI and return the model text output. The default model is Gemini 3.7 Flash High. Pass store=true to receive a conversation_id that can be used with previous_interaction_id on a later call. Returns model text, interaction_id, model, usage metadata and status. Never returns credentials. |
| gemini_modelsA | List the models available to the local Antigravity CLI. Returns filtered model ids and display names. |
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 2 tools
The two tools have clearly distinct purposes: 'gemini' sends prompts and returns text completions, while 'gemini_models' lists available models. There is no ambiguity or overlap between them.
Both names share the 'gemini' prefix, but one is a bare noun and the other uses an underscore suffix. This is readable and predictable, though it doesn't follow a strict verb_noun pattern.
With only two tools, the server feels extremely thin for a general bridge. However, the scope is limited to prompt execution and model listing, so the count is borderline acceptable.
The core functions of sending prompts and listing models are covered, including support for conversation history via IDs. Minor gaps exist, such as streaming or explicit model selection parameters in the main tool, but the essential workflow is usable.