> ## Documentation Index
> Fetch the complete documentation index at: https://docs.neuraltrust.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Overview
> Connect NeuralTrust to AI gateways, coding agents, agent frameworks, applications, and edge platforms
NeuralTrust integrates with the tools already in your AI stack, including
gateways, coding agents, agent frameworks, applications, and edge platforms. Use
this catalog to find the supported connection points and setup instructions for
each tool.
## Products
An integration may support TrustGuard, TrustGate, or both.
| Product | Purpose | Integration capability |
| -------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------- | --------------------------------------------------------------------------------------- |
| **[TrustGuard](/trustguard/overview)** | Runtime security for prompts, responses, tool calls, and tool results. It returns an `allow`, `report`, `transform`, `ask`, or `block` verdict. | Applies policy to traffic at the integration point. The host tool enforces the verdict. |
| **[TrustGate](/trustgate/overview)** | An AI gateway for model provider and MCP traffic. | Centralizes routing, credentials, rate limits, and in-flight payload transformation. |
TrustGuard evaluates individual interactions. TrustGate controls how applications
and agents reach model providers and MCP servers. Some coding-agent plugins
configure both products.
## Find your tool
### AI coding agents
Developer tools and managed coding environments that can read repositories,
execute commands, and call external tools.
TrustGuard lifecycle hooks and TrustGate MCP access are separate controls with
separate credentials. The events available to TrustGuard and the way each verdict
is enforced depend on the coding agent.
| Tool | TrustGuard | TrustGate |
| ---------------------------------------------- | ----------------------------------------------------------------------------------------- | ------------------------------------------------------------------------------ |
| [Claude](/integrations/claude) | Server-side organization screening and local Claude Code hooks | Organization connector for claude.ai, Desktop, Cowork, and Claude Code |
| [Cursor](/integrations/cursor) | A plugin evaluates prompts, tool calls, and tool results | Organization-managed MCP endpoint in place of per-machine server configuration |
| [Codex](/integrations/codex) | Lifecycle hooks evaluate prompts, shell commands, patches, and tool results | TrustGate MCP server configuration for the CLI, desktop app, and IDE extension |
| [GitHub Copilot](/integrations/github-copilot) | Tool calls and tool results in Copilot CLI and VS Code Agent mode; prompts are audit-only | An HTTP MCP server for Copilot Chat |
### AI gateways
Gateway integrations apply policy to model traffic that already passes through
the gateway, without changes to each client application.
Input and output enforcement varies by gateway. Buffered responses can be
evaluated before delivery when the adapter supports it. Streamed output may be
monitor-only after delivery or unsupported, depending on the adapter. Direct
provider calls bypass the gateway.
| Gateway | Integration method |
| -------------------------------------- | ------------------------------------------------------------------------------------------------------------------------ |
| [TrustGate](/integrations/trustgate) | Native integration with no collector API key; covers LLM and MCP traffic; streamed output is monitor-only after delivery |
| [Portkey](/integrations/portkey) | BYOG webhook for Portkey Cloud and self-hosted gateways |
| [LiteLLM](/integrations/litellm) | Custom guardrail file loaded by the LiteLLM proxy |
| [Kong](/integrations/kong) | `ai-custom-guardrail`, on routes running AI Proxy |
| [Apigee](/integrations/apigee) | A Shared Flow, per proxy or per environment |
| [Azure APIM](/integrations/azure-apim) | `send-request` policies |
### Agent frameworks
These integrations connect at the agent, workflow, or tool boundary.
| Framework | Integration method |
| -------------------------------------------------------- | ------------------------------------------------------------------------------------------------------------------- |
| [LangChain](/integrations/langchain) | `langchain-neuraltrust`, an `AgentMiddleware` that hooks the agent loop and can stop a tool call before it executes |
| [n8n](/integrations/n8n) | `@neuraltrust/n8n-nodes-trustguard`, a node whose verdict becomes a branch on the canvas |
| [Microsoft Copilot Studio](/integrations/copilot-studio) | An MCP tool pointed at a TrustGate consumer |
### Edge / WAF
These integrations run in front of AI endpoints you operate. They inspect the
request body on matching routes and return `403` for blocked requests. They do
not inspect responses.
Only traffic that traverses the configured edge route is covered. Request-body
limits, runtime limits, and failure behavior vary by provider.
[Cloudflare](/integrations/cloudflare) ·
[CloudFront](/integrations/aws-cloudfront) ·
[Fastly](/integrations/fastly) ·
[Akamai](/integrations/akamai)
### Build your own
Use an SDK, middleware package, or the REST API for custom applications and
unsupported platforms.
| Integration | Package |
| -------------------------------------------------------------------- | ------------------------------------------ |
| [Python SDK](/integrations/python-sdk) | `trustguard-sdk` |
| [Node.js SDK](/integrations/node-sdk) | `@neuraltrust/trustguard-sdk` |
| [REST](/integrations/rest) | Any HTTP client, any language |
| [Python middleware](/integrations/python-middleware) | FastAPI · Django · Flask |
| [Node.js middleware](/integrations/node-middleware) | Express · Next.js |
| [Go SDK](https://github.com/NeuralTrust/trustguard-sdk/tree/main/go) | `github.com/NeuralTrust/trustguard-sdk/go` |
## Compare coverage
Enforcement depends on the host platform. A gateway that owns the request body
can forward a masked payload, while a provider hook limited to `allow` or `deny`
cannot. The `ask` action also requires an interactive client.
See [Coverage](/integrations/coverage) to compare monitoring, blocking,
redaction, and approval support for LLM and tool-level traffic.
## Employee access to third-party AI
For employee access to AI services your organization does not operate, use the
[browser extension](/platform/browser-extension) on managed Chrome and Edge. The
extension is configured through managed browser policy rather than the collector
catalog.
## How TrustGuard integrations work
Most TrustGuard integrations call the evaluate API. [TrustGate](/integrations/trustgate)
uses a native connection and does not require a collector API key.
```http theme={null}
POST {TRUSTGUARD_URL}/v1/evaluate
Authorization: Bearer
Content-Type: application/json
{
"protocol": "llm",
"direction": "input",
"payload": { "input": "…" },
"session_id": "sess-123",
"consumer_id": "alex@acme.com"
}
```
```text theme={null}
block → deny · ask → apply the host's approval behavior · transform → forward transformed_payload · else → forward
```
Set `direction` to `"input"` before the model call and `"output"` after it. TrustGate
sets the direction automatically. See the [Evaluate API](/trustguard/api/evaluate)
for the full contract.
Create the API key from **Runtime → Collectors → Catalog** in the console. The
selected [collector](/trustguard/concepts/collectors) also provides setup
instructions and a configuration snippet.
Create the policy in **Observe** mode. Observe records decisions in **Activity**
without enforcing them. Review the results, then switch the policy to **Enforce**.
See [Policies](/trustguard/concepts/policies).