Zelinqa MCP
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
| ZELINQA_API_KEY | Yes | Required runtime key, scoped to one published domain | |
| ZELINQA_BASE_URL | No | Base URL for the Zelinqa API. Default: https://api.zelinqa.ai | https://api.zelinqa.ai |
| ZELINQA_SESSION_ID | No | Optional host-owned session to resume after restart | |
| ZELINQA_MAX_RETRIES | No | Retry count. Default: 2 | 2 |
| ZELINQA_CONVERSATION | No | Local name of that resumed session. Default: conversation | conversation |
| ZELINQA_TIMEOUT_SECONDS | No | Per-attempt timeout in seconds. Default: 30 | 30 |
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 |
|---|---|
| zelinqa_startA | Start or recover a named conversation and get its first (or pending) question. Reusing a name returns its current state, not a new session. |
| zelinqa_next_questionA | Report the person's reply to the pending question and get the next one. Without a reply, returns the pending question without consuming a turn. Use exact displayed choice labels; never invent an outcome. For an open question, send the person's words; an outcome alone is refused except refused or asked_no_answer. |
| zelinqa_add_contextB | Report extra conversation context without asking a question or consuming a turn. This text may require language-model analysis. |
| zelinqa_adjustA | Apply confirmed business data, dimension statuses or an objective override without asking a question or consuming a turn. Use configured information and dimension IDs; never invent values or mark an unverified result achieved. |
| zelinqa_statusA | Refresh progress and the pending question. Read-only; also required after an interrupted call or conflict. Reconcile before sending another answer. |
| zelinqa_feedbackB | Record the actual business result: success, partial or failure. This does not itself mark the conversation objective achieved. |
| zelinqa_forgetA | Release a finished conversation from this process's memory. Does NOT delete server data. The host must keep its session ID outside the model to resume later. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
| zelinqa_integration_check | User-selected checklist for testing a Zelinqa agent integration. |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| Zelinqa conversation guide |
TDQS
Scored across 7 tools
Most tools have clear distinct roles (start, status, feedback, forget, next_question), but adjust, add_context and next_question all send information 'without consuming a turn', creating subtle boundaries an agent could misselect between. The detailed descriptions largely resolve this.
All names share a consistent zelinqa_ prefix and are readable, though there is minor mixing of verb forms (adjust, start, forget) with noun forms (status, feedback, next_question). The pattern is predictable enough.
Seven tools is well within the ideal range and each maps to a distinct step in the conversation lifecycle. Nothing feels padded or missing at the count level.
The set covers the full guided-conversation lifecycle: start/recover, advance (next_question), inspect (status), enrich (adjust, add_context), report outcome (feedback), and release (forget). Minor gaps like an explicit conversation-list or end tool are workable around.