CONNECTOR DOCS

Aribot Connector

AI threat modeling, code security review, and compliance automation — available as a connector for Claude, ChatGPT and Gemini. Built by Aristiun / Ayurak.

Overview

Threat Modeling first — Aribot’s core — delivered with full AI Security & Governance across five pillars: Threat Modeling, Code Security (code review, API security, Shadow-AI), Cloud Compliance (cloud security posture), Compliance, and Traceability.

The Aribot connector lets an AI assistant drive your governed Aribot workspace: generate STRIDE threat models, verify threats against real code, trace requirements→controls→remediations, report framework coverage (NIST, ISO 27001, SOC 2), surface remediation guidance, and check live compliance status. Every call is authenticated (OAuth 2.0 + PKCE), scoped to the connecting user's company, and enforced by Aribot's role-based access controls.

Connect

Claude

  1. Settings → Connectors → Add custom connector.
  2. Enter the server URL below; sign in to Aribot and approve consent (Owner or Security Manager).
https://mcp.aribot.ayurak.com/mcp

ChatGPT

Add the same MCP URL (Settings → Apps & Connectors, developer mode), or build a custom GPT importing the OpenAPI schema:

https://api.aribot.ayurak.com/.well-known/openapi.json

Gemini

Register the OpenAPI schema (above) as a tool with OAuth 2.0, or use the remote MCP server.

Claude Code, Gemini CLI, Cursor, VS Code

Local clients sign in through a browser and a callback on your own machine.

claude mcp add --transport http aribot https://mcp.aribot.ayurak.com/mcp

Gemini CLI — in ~/.gemini/settings.json:

{"mcpServers": {"aribot": {"httpUrl": "https://mcp.aribot.ayurak.com/mcp"}}}

Cursor / VS Code — add a remote (HTTP) MCP server with the same URL.

Availability

The Aribot connector is a standards-compliant remote MCP server plus an OpenAPI 3.1 REST surface, so it works across the major AI assistants and their stores. Live means an end user can add it right now; Not listed means Aribot is not in that store's curated directory yet — the connector itself works by adding the URL.

Assistant / storeHow to addStatus
Claude — custom connectorSettings → Connectors → Add custom connector → the MCP URL above.Live
Claude — Connectors DirectoryFind Aribot in Claude's connector directory and connect.Live
ChatGPT — connectorSettings → Connectors → Add → the MCP URL.Live
ChatGPT — GPT Action / GPT StoreImport the OpenAPI schema into a custom GPT (OAuth), then publish.Not listed
ChatGPT — Apps in ChatGPT (Apps SDK)Same MCP server as the app backend.Not listed
Gemini CLI — extensiongemini extensions install https://github.com/Aristiun-B-V/aribot-gemini-extensionLive
Gemini / Vertex AIRegister the OpenAPI schema as an OAuth tool, or add the MCP server.Not listed
MCP Registry (and PulseMCP, Glama, MCP Market)Listed as com.ayurak/aribot-mcp.Live
Claude Code, Gemini CLI, Cursor, VS CodeAdd the remote MCP URL (see Connect).Live

Auth is identical everywhere: OAuth 2.0 + PKCE with dynamic client registration — no per-store secret to manage for the MCP path.

Endpoints

Remote MCP server (Streamable HTTP)  ·  https://mcp.aribot.ayurak.com/mcp
OpenAPI 3.1 schema  ·  https://api.aribot.ayurak.com/.well-known/openapi.json
OAuth authorize / token  ·  https://api.aribot.ayurak.com/o/authorize/  ·  /o/token/
OAuth discovery  ·  https://mcp.aribot.ayurak.com/.well-known/oauth-protected-resource

Tools

The 33 tools the server serves right now (the same list tools/list returns). Read-only tools are marked; the others create or change data in your workspace.

ToolDescriptionAccess
get_billingBilling status + self-service payment for your company: credit-wallet balance, pay-per-use flag, license tier / annual commitment, per-action prices, purchasable plans, and any approved-but-unpaid plans. Pass `checkout_request_id` to get a hosted Stripe Checkout URL to COMPLETE an approved plan, `topup_amount` (EUR) to get one to TOP UP the wallet, or `request_plan` (starter|pay_per_use|pro|max|enterprise) to REQUEST a plan (files a request for super-admin approval — never grants). Returns the details needed to resolve a 402 (licence or credit required).writes
onboard_agentsBulk-onboard agent identities to the governed fleet (Agent Governance). Accepts plain ids or {agent_id} descriptors in `agents`, A2A 1.0 Agent Cards in `agent_cards` (name/url/provider/version/protocolVersion), or MCP client descriptors, under an optional `cohort` + shared auto-suspend policy. Idempotent. Requires the agent_governance licence + a manage:agents grant (or a first-party super-admin). Agents also self-onboard on first token/call.writes
generate_threat_modelCreate a threat model from a normalized architecture (ReactFlow nodes + edges). Aribot stores the components and generates the threats. Returns the diagram id.writes
verify_threats_in_codeVerify whether modelled STRIDE/LINDDUN threats are mitigated in a scan's uploaded code or AST. If threat_id is provided, returns synchronous verdict; otherwise dispatches batch verification across all diagram threats.writes
get_file_requirementsWhat Aribot knows about one source file: the diagram component it is linked to, its security requirements, its threats (with status) and the warning-only IDE gate findings on it. Send the file path and changed line ranges only; file contents are never accepted. Read-only.writes
check_changeCheck changed files against your signed rules with the warning-only IDE gate. Returns findings with file, lines and a fix_ref for get_fix. Nothing is recorded and nothing is blocked. Send paths and changed line ranges only. Read-only.writes
get_fixStored or cached fix steps and suggested patch for one finding, by the fix_ref that check_change or get_file_requirements returned. Never starts a new AI fix; a person can generate one in the Aribot app. Read-only.writes
get_gate_statusLatest gate runs for your company (pull request and build gates): outcome, reasons, finding count and override state. Filter by gate configuration, commit sha or pull request number. At most 20 runs, newest first. Read-only.writes
get_traceabilityReturn the diagram→threat→finding→control→requirement→remediation traceability matrix for a scan in your scope, with coverage metrics.read-only
get_framework_coverageGet control-mapped coverage percentages, gap counts, and mapped controls for EU AI Act, DORA, NIST AI RMF, ISO 27001, SOC 2, etc.read-only
get_remediationCompute a remediation plan for a threat or finding without applying it (dry run). Runs Aribot's governed remediation checks and returns the proposed steps.read-only
compliance_statusCompany-level compliance posture: per-framework posture from the daily cloud scans (checks passed/failed, controls tested/failing), plus threat and control metrics from code scans. Pass `framework` (e.g. 'ISO 27001') to get that framework's most-failing controls. Rates are percentages (0-100).read-only
discover_shadow_aiScan connected repositories and network traffic for unmanaged LLMs, foundational model endpoints, vector databases, agent frameworks, and leaked API keys (async).writes
get_api_securityAPI security inventory (part of Code Security): discovered API endpoints with authentication status, risk level and risk factors, plus method/risk breakdowns. Company-wide or one scan with `scan_id`.read-only
get_cloud_complianceCloud security & compliance posture (Cloud Compliance): per connected cloud account, the latest CIS/NIST cloud-policy scan — daily-scan posture (checks passed/failed) and the last manual scan's compliance %, failing policies/records — plus a company rollup.read-only
code_review_scanStart (or re-run) a code-security scan for an existing scan/repository in your scope. Returns a poll pointer; results include SAST, secrets, deps, pipeline review and the traceability matrix.writes
compliance_scanRun a cloud/platform or compliance scan against an account or diagram in your scope (async). scan_type ∈ platform|compliance|pipeline|sbom. Returns a task id to poll.writes
apply_remediationApply a remediation for real (mode=live). Goes through Aribot's governance checks (attack-path reachability, autonomy policy and the approval flow). If your policy requires approval it returns 'requires_approval' rather than acting.writes
get_diagram_summarySummary of a diagram: threat counts by severity, risk value, compliance and framework coverage.read-only
list_diagramsList your company's threat-model diagrams (id, name, last update), newest first, optionally filtered by part of the name.read-only
get_insightsCode-security metrics for a diagram: the threat → finding → control metrics of its latest code-security scan. code_scan is null when the diagram has no scan yet.read-only
generate_architectureGenerate a multi-tier cloud/AI architecture and auto-synthesize STRIDE/LINDDUN threat models, security requirements, and regulatory control mappings directly from a natural-language description. Runs in the background (a few minutes); returns the name the diagram is saved under.writes
run_ai_governance_auditAudit an architecture diagram or code scan against statutory AI governance frameworks (EU AI Act 2024, EU DORA 2022, NIST AI RMF 1.0, ISO 42001). Evaluates article-by-article conformity, satisfied controls, open gaps, and generates actionable remediation guidance.read-only
get_code_findingsFindings from a code-security scan (SAST, secrets, dependencies): title, severity, file and line, CWE/OWASP category, status and the recommended fix. Latest completed scan when scan_id is omitted.read-only
get_shadow_aiAI usage found in your code and network traffic: provider, service type, model, endpoint, where it was found, risk level, classification and whether a hard-coded key was seen. Company-wide, or one scan with scan_id.read-only
get_agent_governanceYour governed AI-agent fleet: each agent's id, name, cohort, whether it is active, its execution state (running, held, suspended), deviation score and the reason for any hold. Totals by state.read-only
get_self_healingRemediation (self-healing) history: each planned or applied fix with its threat, rule, resource, mode (dry run or live), status and outcome, plus whether autonomous live remediation is switched on for your company.read-only
list_ticket_destinationsTrackers, routes and destinations this caller may send threats to (no credentials).read-only
file_threat_ticketsSend 1-50 threat report entries to the tracker their route points at (or `destination`). One ticket per threat per report; a repeat returns already_filed. dry_run=true returns the exact tickets without sending; a preview with public_repository=true goes to a public repository, visible to anyone. Error codes: permission_missing, role_missing, destination_not_found, already_filed, rate_limited, assignee_unmapped, no_route, no_entry, not_found, invalid, no_company, tracker_error, filed_unrecorded, outcome_unknown, not_attempted, not_configured, tracker_unreachable, unsupported. filed_unrecorded and outcome_unknown mean the ticket may already exist in the tracker; not_attempted means the item was not sent.writes
get_threat_ticketsTickets (tracker, number, state, assignee) for the given threat entries, or for a diagram.read-only
sync_threat_ticketsPull the current state and assignee of this company's open tickets (max 50 per call, time-boxed). Returns {checked, changed, errors, remaining}; remaining > 0 means call again to continue. Error codes: permission_missing, role_missing, destination_not_found, already_filed, rate_limited, assignee_unmapped, no_route, no_entry, not_found, invalid, no_company, tracker_error, filed_unrecorded, outcome_unknown, not_attempted, not_configured, tracker_unreachable, unsupported. filed_unrecorded and outcome_unknown mean the ticket may already exist in the tracker; not_attempted means the item was not sent.writes
assign_threatSet the owner and due date of a threat entry; the linked ticket follows. Error codes: permission_missing, role_missing, destination_not_found, already_filed, rate_limited, assignee_unmapped, no_route, no_entry, not_found, invalid, no_company, tracker_error, filed_unrecorded, outcome_unknown, not_attempted, not_configured, tracker_unreachable, unsupported. filed_unrecorded and outcome_unknown mean the ticket may already exist in the tracker; not_attempted means the item was not sent.writes
comment_threat_on_changeFor a completed pipeline scan of a pull request: post, or update, one comment per threat whose linked code the pull request changes. A re-run with nothing changed makes no calls. Returns each comment's status (posted, updated, skipped with the reason, failed). Public repositories are skipped with the reason 'public repository'. Error codes: permission_missing, role_missing, destination_not_found, already_filed, rate_limited, assignee_unmapped, no_route, no_entry, not_found, invalid, no_company, tracker_error, filed_unrecorded, outcome_unknown, not_attempted, not_configured, tracker_unreachable, unsupported. filed_unrecorded and outcome_unknown mean the ticket may already exist in the tracker; not_attempted means the item was not sent.writes

Authentication & security

ScopeAllows
read:findingsRead code-review findings and traceability
run:codereviewStart code-security scans
read:threatmodelRead threat models
write:threatmodelCreate/upload/generate threat models
run:scanRun platform/cloud scans
read:insightsRead insights, reports and compliance coverage
offline_accessIssue refresh tokens for long-lived assistant connections
read:ticketsRead tracker tickets linked to threats
write:ticketsSend threats to trackers, assign threats, sync ticket status
write:codeComment on pull requests about threats in the code
manage:agentsOnboard AI agents to governance and manage their policies
run:remediationRequest remediations (live changes follow your approval policy)

Data handling

The connector calls Aribot's own first-party API (api.aribot.ayurak.com). It does not collect conversation data beyond what is needed to fulfil a request, and has no access to assistant memory, chat history, or user files. Reads and writes are limited to the authenticated user's company workspace.

Every tool call runs with the signed-in user's own Aribot permissions and is recorded in the workspace audit log. Revoke an assistant's access at any time by removing the connector in the assistant, or in Aribot.

Privacy policy: aribot.ayurak.com/privacy-policy · Terms: aribot.ayurak.com/terms · AI usage: aribot.ayurak.com/legal/ai-usage-policy

Support

Support: support@ayurak.com · Security reports: security@ayurak.com · Privacy: privacy@ayurak.com