JEV Research MCP
Server Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| UI_HOST | No | Dashboard host | |
| UI_PORT | No | Dashboard port | |
| ENABLE_UI | No | Enable local dashboard | |
| LOG_LEVEL | No | Log level; DEBUG for verbose | INFO |
| JEV_API_KEY | No | Alias for TYPESAFE_API_KEY | |
| MAX_SOURCES | No | Server-side cap on max_sources | 10 |
| ENABLE_CACHE | No | Process-local page cache enabled | true |
| BRAVE_API_KEY | Yes | Your Brave Search key (SEARCH_API_KEY fallback) | |
| MAX_REDIRECTS | No | Redirect hops | 5 |
| SEARCH_API_KEY | No | Fallback alias for BRAVE_API_KEY | |
| TYPESAFE_MODEL | No | JEV model | jev-latest |
| SEARCH_PROVIDER | No | Search provider; brave (live only) | brave |
| UI_OPEN_BROWSER | No | Open dashboard in browser automatically | |
| TYPESAFE_API_KEY | Yes | Your TypeSafe key (JEV_API_KEY also accepted) | |
| CACHE_TTL_SECONDS | No | Process-local page cache TTL | 3600 |
| TYPESAFE_BASE_URL | No | API root (gateway-compatible) | https://api.typesafe.ai |
| MAX_DOCUMENT_CHARS | No | Document cap (truncates, never OOMs) | 200000 |
| MAX_RESPONSE_BYTES | No | Streaming cap | 2000000 |
| MAX_SEARCH_RESULTS | No | Search candidates (≤ 50) | 20 |
| JEV_MAX_STATE_CHARS | No | State cap | 80000 |
| MAX_EVIDENCE_TOKENS | No | Evidence budget (approx tokens) | 8000 |
| JEV_BLOCK_BATCH_SIZE | No | Blocks per JEV call | 30 |
| FETCH_TIMEOUT_SECONDS | No | Per-request timeout | 15 |
| JEV_RESULT_BATCH_SIZE | No | Candidates per JEV call | 20 |
| BLOCK_UNCERTAIN_MARGIN | No | Conservative retain band for block filtering | 0.15 |
| MAX_CONCURRENT_FETCHES | No | Fetch semaphore (≤ 20) | 5 |
| MAX_BLOCKS_PER_DOCUMENT | No | Parser cap | 300 |
| RESULT_UNCERTAIN_MARGIN | No | Conservative retain band for result filtering | 0.15 |
| RESULT_USEFUL_THRESHOLD | No | Noul keep threshold for result usefulness | 0.5 |
| BLOCK_RELEVANT_THRESHOLD | No | Block threshold for relevance | 0.5 |
| RESULT_RELEVANT_THRESHOLD | No | Noul keep threshold for result relevance | 0.5 |
| BLOCK_ACTIONABLE_THRESHOLD | No | Block threshold for actionability | 0.5 |
| JEV_MAX_BLOCK_CHARS_FOR_STATE | No | Per-block text sent to JEV (full text kept for evidence) | 2000 |
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
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| researchB | Research a coding question and return compact evidence with provenance. Args: query: The coding/research question (non-empty, max 2000 chars). max_sources: Max sources to fetch (1..server cap, default server-side). max_evidence_tokens: Evidence budget cap (approx tokens, whole blocks only). mode: 'jev' (semantic filtering) or 'baseline' (no JEV, for comparison). |
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
There is only one tool, so there is no possibility of confusing it with another. Its purpose (answer a coding question with sourced evidence) is unambiguous.
With a single tool named 'research', there is no pattern to violate, but it is a bare noun rather than a verb_noun convention. Consistency is trivially satisfied.
One tool for an entire research server is borderline thin; a single rich entry point can work, but there is no separate way to list sources, refine results, or fetch a specific document. It sits at the low end of reasonable scoping.
The tool covers the core research lifecycle (query, sources, evidence, provenance, modes), but notable gaps remain: no way to retrieve or expand a specific source, no follow-up/refinement operations, and no metadata or health operation. Agents can work around these but with friction.