> ## Documentation Index > Fetch the complete documentation index at: https://docs.blindsight.io/llms.txt > Use this file to discover all available pages before exploring further. # Unregister Mcp Tool ## OpenAPI ````yaml /api-reference/openapi.json delete /api/runtime-security/apps/{app_uuid}/mcp/tools/{tool_name} openapi: 3.1.0 info: title: Blindsight API version: 0.1.0 description: >- The full Blindsight REST surface, generated from the running application. Replace the server host with your own deployment. For the Runtime Security integration surface (scan, proxy, tool calls) see the Runtime Security spec, which is hand written and carries worked examples. servers: - url: https://api.your-blindsight.com description: Your Blindsight deployment security: [] paths: /api/runtime-security/apps/{app_uuid}/mcp/tools/{tool_name}: delete: tags: - runtime-security-apps summary: Unregister Mcp Tool operationId: >- unregister_mcp_tool_api_runtime_security_apps__app_uuid__mcp_tools__tool_name__delete parameters: - name: app_uuid in: path required: true schema: type: string title: App Uuid - name: tool_name in: path required: true schema: type: string title: Tool Name responses: '200': description: Successful Response content: application/json: schema: $ref: '#/components/schemas/AppConfigVersionRead' '422': description: Validation Error content: application/json: schema: $ref: '#/components/schemas/HTTPValidationError' components: schemas: AppConfigVersionRead: properties: id: type: string title: Id version_number: type: integer title: Version Number config: $ref: '#/components/schemas/AppConfig-Output' change_summary: anyOf: - type: string - type: 'null' title: Change Summary created_at: type: string format: date-time title: Created At created_by_user_id: anyOf: - type: integer - type: 'null' title: Created By User Id type: object required: - id - version_number - config - created_at title: AppConfigVersionRead HTTPValidationError: properties: detail: items: $ref: '#/components/schemas/ValidationError' type: array title: Detail type: object title: HTTPValidationError AppConfig-Output: properties: thresholds: $ref: '#/components/schemas/AppThresholds' detectors: $ref: '#/components/schemas/AppDetectors' ner: $ref: '#/components/schemas/AppNerPolicy' custom_phrases: items: type: string type: array title: Custom Phrases pii_rules: items: $ref: '#/components/schemas/AppPiiRule' type: array title: Pii Rules tool_policy: $ref: '#/components/schemas/AppToolPolicy' routing: $ref: '#/components/schemas/AppRouting' streaming: $ref: '#/components/schemas/AppStreamingPolicy' egress: $ref: '#/components/schemas/AppEgressPolicy' canaries: $ref: '#/components/schemas/AppCanaryPolicy' redaction: $ref: '#/components/schemas/AppRedactionPolicy' request_signing: $ref: '#/components/schemas/AppRequestSigningPolicy' mcp: $ref: '#/components/schemas/AppMcpPolicy-Output' additionalProperties: false type: object title: AppConfig ValidationError: properties: loc: items: anyOf: - type: string - type: integer type: array title: Location msg: type: string title: Message type: type: string title: Error Type type: object required: - loc - msg - type title: ValidationError AppThresholds: properties: block: type: number maximum: 1 minimum: 0 title: Block default: 0.55 redact: type: number maximum: 1 minimum: 0 title: Redact default: 0.35 output_block: type: number maximum: 1 minimum: 0 title: Output Block default: 0.55 output_redact: type: number maximum: 1 minimum: 0 title: Output Redact default: 0.35 type: object title: AppThresholds description: >- Defaults for a NEW app. Existing apps keep their stored values. 0.55/0.35 rather than 0.85/0.55: measured on 5,514 rows, moving the block line alone takes attack detection from 0.767 to 0.901 and benign pass from 0.937 to 0.919. See DEFAULT_BLOCK_THRESHOLD in the engine for the split validation behind the number. AppDetectors: properties: injection: type: boolean title: Injection default: true pii: type: boolean title: Pii default: true ner: type: boolean title: Ner default: true embedding_anomaly: type: boolean title: Embedding Anomaly default: false perplexity: type: boolean title: Perplexity default: false harm: type: boolean title: Harm default: true topic_drift: type: boolean title: Topic Drift default: false agentic_guardrails: type: boolean title: Agentic Guardrails default: true type: object title: AppDetectors AppNerPolicy: properties: mode: anyOf: - type: string enum: - flag - redact - type: 'null' title: Mode description: >- Unset follows the platform default. 'flag' = record findings and mark the event suspicious without rewriting text; 'redact' = mask NER spans exactly like regex PII hits. threshold: type: number maximum: 0.99 minimum: 0.05 title: Threshold default: 0.55 entities: items: type: string type: array maxItems: 64 title: Entities type: object title: AppNerPolicy description: |- Zero-shot NER for unstructured PII. Enablement lives in ``detectors.ner`` like every other detector; this block only tunes it. Entities are free-form zero-shot labels, adding one requires no retrain. AppPiiRule: properties: name: type: string maxLength: 128 minLength: 1 title: Name pattern: type: string maxLength: 2048 minLength: 1 title: Pattern category: type: string maxLength: 64 minLength: 1 title: Category default: custom score: type: number maximum: 1 minimum: 0 title: Score default: 0.7 description: anyOf: - type: string maxLength: 512 - type: 'null' title: Description flags: type: integer maximum: 258 minimum: 0 title: Flags default: 0 priority: type: integer title: Priority default: 0 type: object required: - name - pattern title: AppPiiRule AppToolPolicy: properties: allowlist: items: type: string type: array title: Allowlist denylist: items: type: string type: array title: Denylist type: object title: AppToolPolicy AppRouting: properties: forbidden_providers: items: type: string type: array title: Forbidden Providers provider: anyOf: - type: string maxLength: 64 - type: 'null' title: Provider description: >- Bound LLM provider slug for the OpenAI-compatible proxy (e.g. 'groq', 'mistral', 'openrouter'). When set, the /openai-compatible/* routes route to this provider by default. upstream_base_url: anyOf: - type: string maxLength: 512 - type: 'null' title: Upstream Base Url description: >- Explicit upstream base URL for the OpenAI-compatible proxy. Must be an https URL allowlisted by the deployment. Takes precedence over 'provider' when both are set. type: object title: AppRouting AppStreamingPolicy: properties: strict: type: boolean title: Strict description: >- When true, every upstream chunk is scanned BEFORE being flushed to the client. Trades TTFB for correctness, closes the gap where small PII/injection payloads escape the inter-scan window. default: false attachment_notice_enabled: type: boolean title: Attachment Notice Enabled description: >- If a request contains an attachment (file, image, document part), inject a short 'analyzing file' notice into streaming responses or set X-Blindsight-Scan-* headers on non-streaming responses. default: true attachment_notice_message: type: string maxLength: 280 title: Attachment Notice Message description: >- Notice text shown inline to the end user when a streaming request contains an attachment. Ignored when attachment_notice_enabled is False. default: '🔒 Blindsight: analyzing attached file for security risks…' type: object title: AppStreamingPolicy description: Per-app streaming behaviour for the proxy. AppEgressPolicy: properties: host_allowlist: items: type: string type: array title: Host Allowlist description: >- Hostnames (exact or *.suffix) the proxy is allowed to call for this app. Empty = inherit deployment allowlist. Always checked AFTER deployment allowlist. type: object title: AppEgressPolicy description: Per-app egress controls layered on top of the deployment defaults. AppCanaryPolicy: properties: enabled: type: boolean title: Enabled default: false phrases: items: type: string type: array maxItems: 8 title: Phrases type: object title: AppCanaryPolicy description: |- Adversarial canary phrases planted in the system prompt. If the model ever echoes one in its output, the response is blocked , this catches jailbreaks the classifier missed. AppRedactionPolicy: properties: style: type: string enum: - bracket - typed_token title: Style description: >- 'bracket' = , ... ; 'typed_token' = , ... (stable per response, parser-friendly for JSON / code outputs). default: bracket type: object title: AppRedactionPolicy description: How PII spans are rewritten in the redacted_text field. AppRequestSigningPolicy: properties: enabled: type: boolean title: Enabled default: false timestamp_skew_seconds: type: integer maximum: 3600 minimum: 30 title: Timestamp Skew Seconds default: 300 type: object title: AppRequestSigningPolicy description: HMAC-signed proxy requests with nonce replay protection. AppMcpPolicy-Output: properties: min_trust_score: type: number maximum: 1 minimum: 0 title: Min Trust Score default: 0.4 tools: items: $ref: '#/components/schemas/McpToolEntry-Output' type: array maxItems: 200 title: Tools enforcement: type: string enum: - advisory - block_high title: Enforcement default: block_high pin_on_first_use: type: boolean title: Pin On First Use default: true scan_results: type: boolean title: Scan Results default: true type: object title: AppMcpPolicy description: MCP tool registry, trust gate, and live verification policy. McpToolEntry-Output: properties: name: type: string maxLength: 128 minLength: 1 title: Name description: type: string maxLength: 8192 title: Description default: '' trust_score: type: number maximum: 1 minimum: 0 title: Trust Score default: 0 severity_top: type: string enum: - none - low - medium - high title: Severity Top default: none findings_count: type: integer minimum: 0 title: Findings Count default: 0 registered_at: anyOf: - type: string - type: 'null' title: Registered At last_scanned_at: anyOf: - type: string - type: 'null' title: Last Scanned At type: object required: - name title: McpToolEntry description: A single MCP tool registered with this app, with its trust score. ````