LLM Gateway Integration
Live

One gateway credential.
Every model on Gyri.

LLM Gateway integration on Gyri routes OpenRouter, LiteLLM, Bifrost, or any OpenAI-compatible API — swap model slugs without losing federated context.

Route OpenRouter, LiteLLM, Bifrost, or any OpenAI-compatible API through llm_gateway_api — swap anthropic/claude-sonnet-4, google/gemini-2.5-pro, and openai/gpt-4.1 without losing federated graph, thought verbs, or MCP write-back.

Start free trial All integrations

Related: Anthropic direct · OpenAI direct · Graph infrastructure

Federation path

LLM Gateway joins
your full stack.

LLM Gateway
Gyri graph
CRM · comms · docs
What connects

LLM Gateway data
in your graph.

Data

BYOK llm_gateway_api credential

Data

Stored Gyri agents (provider: gateway)

Data

OpenRouter, LiteLLM, Bifrost backends

Data

Gemini + multi-vendor model slugs

Data

Embeddings + incorporation via gateway

The problem

LLM Gateway alone
is not enough.

01

Model routing without memory

OpenRouter and LiteLLM aggregate models — but inference is stateless without a company knowledge substrate.

02

Per-model silos

Teams pick Claude for research and GPT for code — each with separate chat history and zero shared institutional recall.

03

Gateway is not the product

Routing APIs do not federate CRM, cite sources, or wish for NL workspace buildout — that is Gyri's job.

Before / after

Gateway alone vs
Gateway + Gyri.

Gateway alone Swap models via HTTP. Every agent rebuilds context. No CRM joins, citations, or recall.
Gateway + Gyri wish for NL buildout. recall prior work. synthesize run-end hubs — any model slug.
With Gyri

Federated, cited,
agent-ready.

Multi-vendor routing

Swap models, keep graph

One gateway credential — change model slugs without rewiring connectors, MCP, or insight persistence.

wish + buildout

NL from any backend

wish and buildout_request queue workspace config regardless of which model powers the agent run.

Embeddings unified

Search shares gateway

Embeddings and agent inference use the same llm_gateway_api — consistent vector + synthesis stack.

Gemini via slugs

google/gemini-* on Gyri

Run Gemini models through gateway slugs with full federated context — no separate Google integration required.

Thinking verbs

Model-agnostic
memory.

Gateway routing picks the inference backend — thought verbs own the epistemic lifecycle.

think verb=wish

Semantic NL intent for connectors and automations

think verb=recall

Same institutional memory regardless of model slug

think verb=ground

Orient on federated CRM + comms before inference

think verb=synthesize

Run-end hubs and promotion candidates

OpenRouter default · LiteLLM self-host · Anthropic direct · OpenAI direct

Competitor matrix

LLM gateway vs LLM gateway + Gyri

Start free trial
Use cases

Built for
GTM operators.

Model failover

Route primary anthropic/claude-sonnet-4 with openai/gpt-4.1 fallback — same recall and promote lifecycle on both.

Self-hosted LiteLLM

Point llm_gateway_api at your LiteLLM proxy — enterprise routing with Gyri federation and MCP.

Cost-aware routing

synthesize run-end hubs with cheaper gateway slugs for bulk tasks — promote only high-confidence insights.

Works with

Federate alongside
native connectors.

Anthropic (Claude)
OpenAI
Cursor Cloud Agents
Notion
GitHub
Setup

Connect in
minutes.

Settings → API keys → LLM Gateway API — set base URL (default OpenRouter) and API key. Optional: GYRI_LLM_GATEWAY_BASE_URL for self-hosted LiteLLM.

1. Start a free trial at app.gyri.io · 2. Open Settings → API keys · 3. Create a stored agent and connect MCP for cited federated answers.

Connect LLM Gateway Setup guide

Connect your LLM gateway to Gyri.

Free trial — OpenRouter, LiteLLM, or custom /v1 endpoint with federated agents.

Start free trial Browse integrations

Frequently asked questions

What LLM gateway backends does Gyri support?

OpenRouter (default), self-hosted LiteLLM, Bifrost, or any OpenAI-compatible /v1 endpoint via llm_gateway_api.

Can I run Gemini through Gyri?

Yes. Use gateway model slugs like google/gemini-2.5-pro — same federated graph and MCP as Claude and GPT agents.

Does the gateway power embeddings too?

Yes. Embeddings and incorporation share llm_gateway_api — consistent vector search and agent inference.