Crow

Overview

Crow is an embeddable AI copilot that plugs directly into your app so users can chat with your product instead of clicking through menus. Upload your OpenAPI spec or connect via MCP, drop in a script tag, and Crow turns your backend endpoints into tools an AI agent can call on behalf of authenticated users. It handles knowledge ingestion from docs, websites, and Notion, while a managed proxy layer enforces authentication, rate limiting, and schema validation.

Key Features

Action Execution via API: Crow doesn't just answer questions. It connects to your backend endpoints, calls your APIs, navigates your UI, and executes real workflows on behalf of authenticated users.
OpenAPI & MCP Integration: Wire in your app's functionality by uploading an OpenAPI spec or connecting an MCP server. Crow auto-generates tools from your endpoints so the agent knows what it can do.
Knowledge Base Ingestion: Feed Crow your docs, files, website content, or Notion pages. The agent uses this context to give accurate, product-aware answers.
Managed Security Proxy:Security and authentication layer that verifies users and ensures the agent only performs allowed actions.
Custom Workflows & Guardrails: Build workflows for specific scenarios and set guardrails so the agent operates within boundaries you define.
Simple Embedding Options: Add Crow with a single script tag or use the React SDK (@usecrow/ui) for a floating widget or copilot sidebar. No complex setup required.
Conversation Monitoring: Track every conversation, tool call, and action through the Activity dashboard. See what users ask for and where things break.
BYO LLM Support: Supports flexible model configuration, including routing requests through your preferred LLM provider.

Pros

Fast time to production: Customers report going live in under a week, far quicker than building an in-house AI copilot from scratch.
Takes real actions, not just chat: Unlike standard chatbots, Crow actually executes API calls and navigates your product's UI on behalf of users.
Strong security model: Strong permission and identity controls with JWT verification, tool allow-lists, and validated API execution.
Clean, minimal integration effort: A single script tag or React component is all you need to embed it. The OpenAPI spec import auto-generates tools.
Full data control: Strong data control options, including choosing what context is sent to the model and support for routing requests through your own LLM provider.

Cons

Requires API-first architecture: Crow works best with apps that have a well-documented OpenAPI spec or MCP server. Products without clean API layers will need extra preparation.

Use Cases

ERP and Workflow Software: Users of complex enterprise apps can type natural language commands like "create a new customer" or "open the P&L page" instead of navigating deep menu structures. Crow executes the underlying API calls, reducing training time and support tickets.
SaaS Customer Self-Service: For products with large user bases, Crow handles common requests like "cancel this subscription," "change my plan," or "export last month's data" directly through chat. This reduces the load on support teams while giving users instant results.
Developer Platforms and Consoles: Users of cloud consoles or API management tools can issue commands like "rotate this API key" or "add a new user" through Crow instead of hunting through settings pages. The agent maps these to the right API calls automatically.
Insurance and Financial Services Apps: Financial or account-based applications where users need quick access to information or actions without navigating complex portals.
Making Legacy Software AI-Native: Teams with existing workflow-heavy products can add a chat-first interface without rebuilding their frontend. Crow sits on top of your current app and brings a modern AI copilot experience to users within days.

Frequently Asked Questions

What is Crow?

Crow is an embeddable AI agent that connects to your product's APIs, knowledge base, and UI. Instead of users clicking through menus, they type what they want in a chat interface and Crow executes the actions for them.

How does Crow differ from a regular chatbot?

Most chatbots only answer questions. Crow goes further by connecting to your backend endpoints and executing real actions like creating records, navigating pages, and triggering workflows on behalf of authenticated users.

How do I integrate Crow into my product?

The integration process has four steps: embed a script tag or React component, configure the agent's instructions and knowledge in the Sandbox, connect your API via an OpenAPI spec or MCP server, then deploy and monitor conversations through the Activity dashboard.

Is Crow secure?

Crow uses your existing authentication (JWT or session tokens) plus a shared secret to verify every request. A managed proxy handles rate limiting, schema validation, and request logging. You control which endpoints the agent can access through a strict tool allow-list, and Crow does not store customer data.

Can I use my own LLM with Crow?

Yes. Crow supports routing to your own LLM provider, so you maintain full control over which models process your data. This is useful for organizations with strict data privacy requirements.

How long does it take to deploy Crow?

According to the Crow team, customers typically go live in under a week. The setup involves uploading your API spec, configuring the knowledge base, and embedding a script tag or React component.

How much does Crow cost?

Crow offers a free option for developers with side projects. For production use, pricing is not publicly listed. You'll need to contact the Crow team directly to discuss plans tailored to your usage and scale.