mcp-vaara
OfficialServer Configuration
Describes the environment variables required to run the server.
| Name | Required | Description | Default |
|---|---|---|---|
No arguments | |||
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
} |
| resources | {
"subscribe": false,
"listChanged": false
} |
Tools
Functions exposed to the LLM to take actions
| Name | Description |
|---|---|
| vaara_checkA | Check the risk level of an action BEFORE executing it. Returns a risk assessment with point estimate, conformal interval, and recommended decision (allow/deny/escalate). Does NOT record to the audit trail — use vaara_intercept for that. |
| vaara_interceptA | Intercept an action with full risk scoring and audit trail. Classifies the action, scores risk with conformal prediction, decides allow/deny/escalate, and records everything in a tamper-evident audit trail. Returns the decision and action_id for later outcome reporting. |
| vaara_report_outcomeA | Report what happened after an action was executed. This closes the feedback loop — the scorer learns from outcomes to improve future risk assessments. Provide the action_id from vaara_intercept and the observed severity. |
| vaara_verify_receiptA | Verify a signed decision receipt: its signature, and that the evidence it carries is the evidence it signed. For a receipt under this server's own key it also checks that the record it names is in this server's audit trail. Pass the receipt file's JSON as 'receipt'. The server's key is used unless 'public_key_pem' names another issuer's key; then the trail is not checked. |
Prompts
Interactive templates invoked by user choice
| Name | Description |
|---|---|
No prompts | |
Resources
Contextual data attached and managed by the client
| Name | Description |
|---|---|
| Vaara Status | Current scorer status: calibration state, MWU weights, thresholds, tracked agents |
| Vaara Compliance Report | Latest EU AI Act, DORA and SOC 2 compliance assessment against the audit trail |
TDQS
Scored across 4 tools
vaara_check and vaara_intercept overlap substantially—both classify, score risk with conformal prediction, and return allow/deny/escalate. The descriptions explicitly differentiate them (check = no audit recording; intercept = full audit trail), which resolves most confusion, but the boundary still requires careful reading.
All four tools follow a strict 'vaara_' prefix plus clear verb_noun pattern (check, intercept, report_outcome, verify_receipt). The convention is predictable and readable throughout.
Four tools is a tight, well-scoped set for a risk-scoring and governance service, and each has a distinct role in the decision lifecycle. Slightly thin but no tool feels redundant or filler.
The set covers scoring, interception with audit recording, outcome feedback, and receipt verification, forming a closed loop. A read/list tool for the audit trail itself is absent, a minor gap agents could work around via verify_receipt.