Open Jev Bridge
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| SYSTEM_ONE_URL | No | The base URL of the System One backend. | http://127.0.0.1:8009 |
| SYSTEM_ONE_MODEL | No | The model identifier to use. | kev-latest |
| SYSTEM_ONE_API_KEY | No | Bearer token for authenticated backends (e.g., hosted Jev). Optional unless required by the backend. | |
| SYSTEM_ONE_PROVIDER | No | Provider profile: generic, jev, kev, or laya. | generic |
| OPEN_JEV_BRIDGE_DATA | No | Path to the data/state directory. | |
| SYSTEM_ONE_TIMEOUT_MS | No | Request timeout in milliseconds. | 20000 |
| OPEN_JEV_BRIDGE_CONFIG | No | Path to a JSON configuration file overriding defaults. | |
| SYSTEM_ONE_AUTO_REVIEW | No | Automatic patch/completion gate after fresh passing checks. | true |
| SYSTEM_ONE_AUTO_SCREEN | No | Automatic bounded external-result screening. | true |
| SYSTEM_ONE_AUTO_VERIFY | No | Automatic completion/subagent/task checks. | true |
| SYSTEM_ONE_JEV_ALIASES | No | Set to '1' to enable jev_* compatibility aliases. | 0 |
| SYSTEM_ONE_ALLOW_REMOTE | No | Set to '1' to allow non-loopback URLs. | 0 |
| SYSTEM_ONE_API_KEY_FILE | No | Path to a private file containing the API key. Alternative to SYSTEM_ONE_API_KEY. | |
| SYSTEM_ONE_BELAY_SHADOW | No | Set to '1' for shadow mode (logs would-block without enforcing). | 0 |
| SYSTEM_ONE_AUTO_COMPACTION | No | Bridge checkpoint/function compaction. | true |
| SYSTEM_ONE_COMPACT_AT_PERCENT | No | Optional Claude function usage trigger for compaction. | 60 |
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 |
|---|---|
| system_one_verifyA | Verify claims against supplied evidence; probability-based advisory verdicts, not proof. |
| system_one_screenB | Screen supplied external text for prompt injection, relevance and substance. Never executes or fetches the text. |
| system_one_findA | Find which supplied candidates answer a query without embeddings; return ranked candidates and an existence judgment. |
| system_one_classifyB | Classify supplied items against a shared class catalog; ambiguous results require review. |
| system_one_decideB | Compare 2–6 candidates with explicit requirements and escape hatches. Advisory only; does not act on the decision. |
| system_one_rerankA | Independently score relevance of each supplied candidate, then sort; candidate order breaks ties. |
| system_one_compareB | Compare two passages overall and by aspect. Agreement is not evidence that either passage is true. |
| system_one_extractB | Find bounded regex candidates in a worker, then select a verbatim source substring. Never generates extracted values. |
| system_one_reviewA | Score a proposed diff for correctness, spec match, test gap, blast radius and safety. Does not execute tests or apply patches. |
| system_one_gateB | Review a diff and verify completion claims together in one bounded System One request. Does not execute tests. |
| system_one_queryD | Validated raw System One request; state is arbitrary JSON. This is not an OpenAI chat-completions endpoint. |
| system_one_compactB | Verbatim tool-pair compaction of supplied canonical messages; returns a new transcript and never modifies host files. |
| system_one_belayB | Judge whether a supplied transcript makes unsupported completion claims. Returns an advisory Stop-hook decision; runs no commands. |
| system_one_statusA | Check the configured System One /v1/models endpoint and report bridge capabilities; exposes no API credentials. |
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 14 tools
Each tool has a clearly distinct purpose: from checking claims (belay) to ranking candidates (rerank) to validating queries (query). While some overlap in concept (compare vs decide vs verify), the descriptions clearly delineate their specific roles and inputs, leaving no ambiguity about which tool to use.
All tools follow the consistent pattern 'system_one_' + a verb in lowercase snake_case (e.g., system_one_belay, system_one_compare). The prefix groups them coherently, and the verb accurately reflects each action, making the naming predictable and easy to navigate.
With 14 tools, the server is well-scoped for a specialized verification and analysis system. Each tool addresses a distinct aspect of the domain (e.g., extraction, classification, reranking, review), and none feel redundant or extraneous. The count fits comfortably within the ideal range.
The tool surface covers a comprehensive set of operations for a safety-focused analysis server: status checking, raw queries, extraction, comparison, verification, classification, ranking, screening, review, and gate decisions. There are no obvious missing lifecycle operations, and the domain is not one that requires CRUD-style updates or deletions.