# Blekline Docs (llms-full.txt) # https://app.blekline.com/docs NHIM and runtime enforcement for production AI agents — mask secrets, enforce tool policy, block contaminated lineage, metadata-only audit at the MCP, SDK, and ingress sidecar boundary. Marketing index: https://blekline.com/llms.txt OSS: https://github.com/Blekline/blekline-oss Last-Updated: 2026-08-20 Sidecar image: ghcr.io/blekline/sidecar:0.2.1-nhim NHIM audit: @blekline/nhim-audit@0.2.1 ## Feeds (machine-readable) - Threats RSS: https://app.blekline.com/marketing/threats/feed.rss - Threats catalog: https://app.blekline.com/marketing/threats/catalog.json - Arena scores: https://app.blekline.com/marketing/arena/latest.json - Benchmarks: https://app.blekline.com/marketing/benchmarks/latest.json - Marketing sitemap: https://blekline.com/sitemap.xml ## Get started - /docs — Quick Start — connect Blekline in minutes (Claude Code, Cursor, Claude Desktop). - /docs/guides — Integration guides hub — Claude, Cursor, VS Code, agent stacks. - /docs/reference — API & SDK reference hub. - /docs/platform — Enterprise platform hub — SSO, SCIM, in-VPC, procurement. - /docs/enterprise/implementation-checklist — Enterprise implementation checklist — Audit-first in-VPC pilot, dual-control enforce, committee evidence. - /docs/enterprise/cursor-verification — Cursor DPO verification — 10-step retest and guarantee matrix for IDE governance. - /docs/get-started/eval-journey — Eval journey — Choose your deployment track — Kubernetes fleet, Docker sidecar, or MCP — and prove runtime enforcement on your stack in hours, not quarters. - /docs/get-started/nhim-audit-quickstart — NHIM audit quickstart — Scan your Kubernetes cluster for agent workloads that bypass runtime enforcement — static NHIM posture in minutes, no Blekline account required. - /docs/get-started/mcp-eval — MCP self-serve eval — Mask prompts and enforce MCP tool calls in Cursor, Claude Code, or VS Code — cloud policy live in under 30 minutes, no platform eval required. ## Core concepts - /docs/introduction/interaction-governance — Runtime enforcement — Non-Human Identity and Runtime Enforcement at Layer 4 — mask ingress, enforce execution lineage, redact egress, and stream metadata-only audit to your SIEM. - /docs/introduction/nhim — NHIM overview — Non-Human Identity Management and AI runtime enforcement at the agent boundary. - /docs/introduction/why-ingress — Why Ingress — Why Layer 4 governance is required before autonomous agents scale. - /docs/introduction/egress-governance — Egress governance — Response masking and outbound API governance for autonomous agents at runtime. - /docs/introduction/architecture — Architecture — Blekline at Layer 4; five layers from infrastructure to autonomous agents. - /docs/introduction/troubleshooting — Troubleshooting — Common Blekline setup questions, MCP errors, and local vs cloud modes. ## Deploy - /docs/deploy/docker-sidecar — Docker sidecar — Track 02 platform eval — Trust Vault, Lineage Firewall, and ingress proxy on ghcr.io/blekline/sidecar. Validate VPC enforcement in ~1 hour without a full cluster rollout. - /docs/deploy/k8s-fleet — Kubernetes fleet — Track 01 platform eval — mandatory-hop sidecar, Trust Vault, Lineage Firewall, and fleet policy in your cluster. Prove production-grade NHIM enforcement in ~2 hours. - /docs/deploy/ci-nhim-gate — CI NHIM gate — Merge-blocking NHIM cluster audit in GitHub Actions and CI pipelines. - /docs/tools/nhim-audit — NHIM Audit CLI — Static Kubernetes audit for agent execution path bypass — evidence enablement, not certification. ## MCP clients - /docs/mcp/cursor — Cursor — MCP setup and chat auto-mask hooks for all Cursor models and modes. - /docs/mcp/claude-code — Claude Code — MCP setup for Claude Code via project settings. - /docs/mcp/claude-desktop — Claude Desktop — P1 — native Claude app MCP integration (manual + OAuth connector). - /docs/mcp/claude-connector — Claude connector — OAuth MCP connector for Claude Desktop and Claude.ai remote MCP. - /docs/mcp/codex — Codex — P2 — Codex app TOML MCP configuration. - /docs/mcp/continue — Continue — P3 — MCP setup for Blekline ingress governance. - /docs/mcp/github-copilot — GitHub Copilot — P4 — MCP setup for Blekline ingress governance. - /docs/mcp/openhands — OpenHands — P3 — MCP setup for Blekline ingress governance. - /docs/mcp/sourcegraph-cody — Sourcegraph Cody — P3 — MCP setup for Blekline ingress governance. - /docs/mcp/server — MCP Server — blekline-mcp-server tools and env vars. - /docs/mcp/proxy — MCP Proxy — Tool-call governance before downstream sandbox execution. - /docs/mcp/model-matrix — Client QA Matrix — Manual QA checklist for MCP tools across agent clients and models. ## SDK - /docs/sdk/typescript — TypeScript SDK — @blekline/client REST wrapper. - /docs/sdk/python — Python SDK — blekline-client — Python wrapper for the Blekline REST API. - /docs/sdk/ci-cd — CI/CD — Gate agent tool calls in GitHub Actions and pipelines. ## Integrations - /docs/integrations/agent-clients — Agent Clients — Wire Blekline into Cursor, Claude Code, GitHub Copilot, VS Code, and every MCP surface your team already uses — mask, enforce, and audit before tools run. - /docs/integrations/model-providers — Model Providers — Route OpenAI-compatible, Anthropic, and multi-model traffic through Blekline ingress — mask prompts and enforce policy before requests reach Azure, Bedrock, Vertex, or any router. - /docs/integrations/frameworks-and-rag — Frameworks & RAG — L2 agent frameworks and vector stores with Blekline governance. - /docs/integrations/eval-and-safety — Eval & Safety — L3 eval and post-model safety integrations complementary to Blekline. - /docs/integrations/azure-openai-stack — Azure OpenAI Stack — Layer 4 Blekline governance + Azure OpenAI models. - /docs/integrations/aws-bedrock-stack — AWS Bedrock Stack — Layer 4 Blekline governance + AWS Bedrock models. - /docs/integrations/google-vertex-stack — Google Vertex Stack — Layer 4 Blekline governance + Google Vertex AI models. - /docs/integrations/openrouter-stack — OpenRouter Stack — Layer 4 Blekline governance + OpenRouter models. - /docs/integrations/together-stack — Together Stack — Layer 4 Blekline governance + Together AI models. - /docs/integrations/fireworks-stack — Fireworks Stack — Layer 4 Blekline governance + Fireworks AI models. - /docs/integrations/mistral-stack — Mistral Stack — Layer 4 Blekline governance + Mistral AI models. - /docs/integrations/cohere-stack — Cohere Stack — Layer 4 Blekline governance + Cohere models. - /docs/integrations/replicate-stack — Replicate Stack — Layer 4 Blekline governance + Replicate models. - /docs/integrations/qwen-stack — Qwen Stack — Layer 4 Blekline governance + Qwen models. - /docs/integrations/langchain-stack — LangChain Stack — Blekline ingress governance with LangChain. - /docs/integrations/llamaindex-stack — LlamaIndex Stack — Blekline ingress governance with LlamaIndex. - /docs/integrations/pydantic-ai-stack — PydanticAI Stack — Blekline ingress governance with PydanticAI. - /docs/integrations/taskweaver-stack — TaskWeaver Stack — Blekline ingress governance with TaskWeaver. - /docs/integrations/relevance-ai-stack — Relevance AI Stack — Blekline ingress governance with Relevance AI. - /docs/integrations/supabase-stack — Supabase Stack — Blekline ingress governance with Supabase pgvector. - /docs/integrations/chroma-stack — Chroma Stack — Blekline ingress governance with Chroma. - /docs/integrations/pinecone-stack — Pinecone Stack — Blekline ingress governance with Pinecone. - /docs/integrations/contextual-ai-stack — Contextual AI Stack — Blekline ingress governance with Contextual AI. - /docs/integrations/langsmith-stack — LangSmith Stack — Blekline ingress audit metadata with LangSmith eval traces. - /docs/integrations/guardrails-stack — Guardrails Stack — Complementary pre-model Blekline and post-model Guardrails validation. - /docs/integrations/coderabbit — CodeRabbit Workflow — Blekline-governed dev agents alongside CodeRabbit PR review. - /docs/integrations/sandbox-providers — Sandbox Providers — Compare L1 sandbox runtimes for Blekline MCP proxy wiring. - /docs/integrations/daytona-stack — Daytona Stack — Layer 4 Blekline governance + Layer 1 Daytona runtime. - /docs/integrations/modal-stack — Modal Stack — Layer 4 Blekline governance + Layer 1 Modal serverless sandboxes. - /docs/integrations/vercel-sandbox-stack — Vercel Sandbox Stack — Layer 4 Blekline governance + Layer 1 Vercel Sandbox microVMs. - /docs/integrations/cloudflare-stack — Cloudflare Stack — Layer 4 Blekline governance + Layer 1 Cloudflare Containers and Sandbox SDK. - /docs/integrations/e2b-stack — E2B Stack — Layer 4 Blekline governance + Layer 1 E2B Firecracker sandboxes. - /docs/integrations/browser-extension — Browser Extension — Blekline extension as browser ingress wedge for ChatGPT, Claude, Gemini, and more. ## Enterprise - /docs/mcp/cursor — Cursor — Enforced hook layers, guarantee matrix, IDE vs cloud agent coverage, and production ingress for DPO review. - /docs/enterprise/deployment — Deployment — Private tenant, MDM extension rollout, and MCP fleet configuration. - /docs/enterprise/compliance-evidence — Compliance evidence — How Blekline helps collect OWASP ASI and AIUC-1 audit evidence — not certification. - /docs/enterprise/k8s-deployment — Kubernetes deployment — Standalone sidecar Helm chart and mutating admission auto-inject for NHIM runtime enforcement. - /docs/enterprise/reference-architecture — Reference architecture — Kong + Blekline + Okta complement — Layer 3 gateway, Layer 4 agent enforcement, Layer 2 human IAM. - /docs/enterprise/shadow-mcp-detection — Shadow MCP detection — Discover unsanctioned MCP servers from Activity telemetry and nhim-audit agent candidates. - /docs/enterprise/trust-vault-sidecar — Trust Vault sidecar — Deploy stateful secret tokenization in your VPC sidecar. - /docs/enterprise/lineage-enforcement — Lineage enforcement — Block destructive tools when agent session lineage is contaminated. - /docs/enterprise/nhim-verification — NHIM verification — Public verification checklist for NHIM sidecar and Kubernetes deployments. - /docs/enterprise/multi-region — Multi-region Ingress — Deploy Blekline edge sidecars near agents in five regions alongside Daytona runtime. - /docs/enterprise/sso-deployment — SSO & Deployment — Enterprise SSO roadmap, deployment models, and hardening checklist. - /docs/enterprise/telemetry — Telemetry — Fleet sensor metadata, event ingest fields, and observability boundaries for Blekline governance — metadata-only by default. - /docs/enterprise/procurement-pack — Procurement Pack — Security, privacy, and OSS boundary summary for enterprise procurement. ## API - /docs/api/mask — POST /api/mask — Authoritative masking gateway. - /docs/api/openapi — OpenAPI — Blekline OpenAPI 3.1 spec — mask, events, MCP enforce, ingress proxy, and workspace policy endpoints. - /docs/api/ingress-proxy — Ingress Proxy — OpenAI and Anthropic compatible routes with inline masking before upstream model APIs. ## Reference - /docs/reference/telemetry-spec — Telemetry Spec — Formal event metadata schema for fleet sensors (whitepaper §7.4.3 alignment). - /docs/reference/latency-slo — Latency SLO — Enterprise ingress latency targets and measurement. - /docs/reference/benchmarks — Benchmarks — Lab-tested agent-boundary benchmarks — reproducible methodology and raw JSON artifacts. ## Architecture & security - /docs/security/trust-boundaries — Trust Boundaries — Data handling, persistence, and security posture for enterprise diligence. - /docs/security/threat-model — Threat Model — Blekline ingress, sidecar, and extension threat model for security review. - /docs/security/mcp-identity-pinning — MCP Identity Pinning — Pin MCP server command paths to prevent config swap attacks (MCPoison-style). ## Playground - /docs/playground/runtime-enforcement — Runtime Simulator — Interactive NHIM simulator — NHI attestation, YAML policy, tool payloads, and SIEM export. ## Definitions - /docs/definitions — Glossary