OpenRouter

Reach hundreds of models from OpenAI, Anthropic, Google, Meta, Mistral, and DeepSeek through one unified OpenRouter API, with automatic routing, fallbacks, and cost insights.

18 actions available

New task queued that needs a model call

Agent sends the request through OpenRouter with a preferred model

OpenRouter falls back to an alternate model if the first is unavailable

Agent reads the structured response

Agent records the result and the model that served it

The owning team gets the output and cost insight

A result the model is unsure about routes to a person to decide

What This Integration Enables

OpenRouter is a single door to the whole model market. It covers chat, images, audio, video, and embeddings across hundreds of models, routes requests automatically, falls back when a provider fails, and reports cost and usage across everything. Swapping models or A/B testing them becomes a parameter change, not a new integration. The resilience is the operational win: an AI step that can fail over to another provider does not take the whole flow down when one vendor has a bad day. That reliability, plus cost visibility across models, makes a flow's AI layer dependable. An orchestration layer routes the output and inserts the human gate, and FlowRunner is built for that layer.

Without FlowRunner

One provider, one point of failure A vendor outage stalls every AI step

Per-vendor integrations Each new model means a new integration to build

No cost visibility Model spend scattered across separate vendor bills

With FlowRunner

Automatic fallback A request reroutes to another model on an outage

Every model, one key Hundreds of models behind a single interface

Cost insight built in Spend and usage visible across all models

Use Case Scenarios

Resilient AI Steps

A production flow depends on a model call that cannot afford to fail. The agent sends the request through OpenRouter with a preferred model and an automatic fallback. If the first provider is down or slow, the request reroutes to an alternate and the flow keeps running. The AI step stops being a single point of failure, and the result records which model served it.

Model Comparison in Production

A team wants to find the best model for a task on real traffic without rebuilding an integration each time. The agent routes a share of requests to different models through OpenRouter and records the outputs and costs. A person reviews the comparison and sets the default, turning model selection into a measured decision.

Cost-Aware Routing

Different tasks justify different models. The agent routes simple, high-volume requests to a cheaper model and complex ones to a stronger model, all through OpenRouter, and uses the cost insights to keep spend in view. The uncertain results still route to a person, so cost optimization never trades away the human check.

Human-in-Loop Highlight

Reaching every model does not change who owns the decision behind the output. When a result comes back with low confidence, or drives an action with consequences, FlowRunner routes it through a [human-in-the-loop](/concepts/human-in-the-loop/) step regardless of which model served it: the agent pauses, presents the output and its context, and sends it to the right person via Slack. They decide. OpenRouter keeps the AI layer flexible and resilient; a person owns the calls that matter.

Agent processes routinely

Detects exception requiring judgment

Clear match Continues automatically

Ambiguous Routes to human via Slack

Human decides

Agent resumes with decision

Agent Capabilities

18 actions

Chat

Chat Completion Generates a text response for a single prompt using any of the hundreds of models available through OpenRouter (OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek and more) via one unified API.
Chat Completion (Advanced) Sends a fully custom chat completion request through OpenRouter's unified API with complete control over the request: full messages array (multi-turn conversations, multimodal content parts), model fallbacks, the complete provider routing preferences object (order, only, ignore, quantizations, sort, allow_fallbacks, data_collection, max_price, require_parameters), reasoning configuration, plugins (web search, file parser, auto router, context compression and more), tool/function calling passthrough, structured outputs via a response format object (json_object or json_schema) and all sampling parameters.
Analyze Image Analyzes one or more images with any multimodal model available through OpenRouter and answers a prompt about them, describe content, extract text, compare images or answer visual questions.

Images

Generate Image Generates one or more images from a text prompt using any image generation model available through OpenRouter (Seedream, FLUX, Gemini image models, GPT Image and more).

Audio

Text to Speech Converts text into speech audio using any text-to-speech model available through OpenRouter (e.g. OpenAI TTS, Voxtral).
Transcribe Audio Transcribes an audio file into text using any speech-to-text model available through OpenRouter (Whisper, Voxtral, Gemini and more).

Video

Generate Video Submits an asynchronous video generation job using any video model available through OpenRouter (Veo, Sora, Seedance and more).
Get Video Status Polls the status of a video generation job by its job ID. The status progresses through pending, in_progress and completed (or failed/cancelled/expired).
Download Video Downloads the finished video content of a completed video generation job, saves it to FlowRunner file storage and returns its URL.

Embeddings

Create Embeddings Generates embedding vectors for one or more texts using any embeddings model available through OpenRouter (OpenAI, Google, Qwen, Mistral and more).
Rerank Documents Reranks a list of documents by relevance to a search query using a rerank model available through OpenRouter (e.g. Cohere Rerank).

Insights

Get Generation Retrieves detailed request and usage metadata for a past generation by its ID (returned by every chat completion), including the exact total cost in credits, native and normalized token counts, cached/reasoning token counts, provider name, latency, generation time and finish reason.

Models

List Models Lists models available through OpenRouter with rich filtering: free-text search, use-case category, input/output modalities, minimum context length, maximum prompt price and server-side sorting (popularity, newest, price, context, throughput, latency, intelligence).
Get Model Endpoints Retrieves all provider endpoints serving a specific model, with each endpoint's provider name, pricing, context length, max completion tokens, quantization, supported parameters, uptime and status.
List Providers Lists all inference providers available through OpenRouter with their slugs (for use in provider routing preferences), headquarters, datacenter regions, and privacy policy, terms of service and status page URLs.

Account

Get Credits Retrieves the total credits purchased and total credits used for the OpenRouter account. The remaining balance is total_credits minus total_usage, useful for balance monitoring and low-credit alerts in flows.
Get Key Info Retrieves information about the API key used for this connection, including its label, spending limit, remaining limit, limit reset interval, free-tier status and current usage, useful for monitoring spend and rate limit headroom in flows.
Get Activity Retrieves the account's usage activity for the last 30 days, grouped by day, model and provider, each item includes request count, prompt/completion/reasoning token totals and spend.

Start building with OpenRouter

$100 in credits. No card required. Connect in minutes.

Start Building Free Talk to us