alibi
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| TYPESAFE_API_KEY | Yes | API key for TypeSafe AI's Jev model, required to use the judge backend. | |
| ALIBI_JUDGE_BACKEND | No | Backend for the judge, typically 'typesafe'. | |
| ALIBI_MAX_TRACE_TOKENS | No | Maximum trace tokens allowed (default: 250000). | |
| ALIBI_MIN_TRACE_TOKENS | No | Minimum trace tokens for analysis (default: 50000). | |
| ALIBI_ALLOW_PAID_MODELS | No | Set to '1' to allow paid model calls; without it, the judge refuses to run. |
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
} |
| prompts | {
"listChanged": false
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
| experimental | {} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| diagnose_traceA | Find where a long AI-agent run went wrong. Reads the trace chapter by chapter
(never all at once), raises a CUSUM alarm, looks back, and returns the 3 steps to
read first. |
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 1 tool
With only one tool, there is zero possibility of confusion or misselection. The tool's purpose is clearly distinct by virtue of being the sole tool in the server.
The single tool name 'diagnose_trace' follows a clear verb_noun pattern and accurately describes its action and target. There is no inconsistent naming to flag.
A single tool feels thin, but the server's apparent scope is narrowly focused on trace diagnosis. The tool consolidates several internal steps (reading, alarming, looking back, returning recommendations), making one tool borderline acceptable.
For its stated purpose, the tool covers the full diagnostic workflow in one call. Minor gaps exist (e.g., no configurable gates for the length/cost thresholds, no follow-up actions), but agents can achieve the core outcome without dead ends.