The operating system for your AI agents.
One platform to build, operate and measure agents, from the first prototype to the operation in production, with observability, cost control and enterprise compliance. In code, through the SDK, or by talking, in the Studio.
Build, operate and measure. On the same runtime.
The agent's entire lifecycle in one place, from the first prototype to the business indicator: with governance running through every stage, not bolted on at the end.
- TypeScript SDK
- Conversational Studio
- Prompt Studio
- Knowledge base (RAG)
- Templates and library
- Automatic tracing
- Debug and replay
- Multi-LLM with fallback
- 150+ connectors and MCP
- Isolated staging and production
- Business dashboards
- Real-time events
- Cost per execution
- Performance review
- Execution reviews
What separates an operation from an experiment.
Three capabilities separate companies that operate AI from companies that only run pilots. The platform is built on them.
Governance
Guardrails, access control, human-in-the-loop and cost visibility by team and by user. The agent acts inside your rules.
Traceability
Every execution, decision and tool call recorded and reproducible. You audit the past and debug the present.
Speed
From prototype to production in weeks. SDK, Studio and templates that take the agent off the slide and put it to work.
Two ways to build. Zero lock-in.
The technical team builds in code, through the SDK. The business team builds by talking, in the Studio, just below. The same agent, the same runtime.
TypeScript SDK
@runflow-ai/sdk, type-safe end to end. Define agents, tools, memory and RAG in code versioned in your Git, with a supervisor pattern for multi-agent and workflows.
- Tools with Zod
- Memory + auto-summary
- Agentic RAG
- Fluent workflows
- Multi-agent supervisor
- Multi-LLM
Describe the operation. The Studio assembles the agent.
The browser environment for people who do not write code. The business team describes what the agent needs to do, and the Studio generates the architecture, connects the integrations and opens an agent ready to iterate. The description becomes a working prototype in minutes, on the same runtime as the SDK.
Build with the Studio
Describe the agent and the Studio assembles the structure, then you iterate by talking to it.
See every step. Debug any execution.
Automatic observability in production, with nothing to instrument. Every agent and workflow generates detailed traces, with cost per token and replay of any past decision.
Business metrics, not only technical ones.
Emit business events from the agent and assemble the dashboard without a separate BI tool: cards, series and funnels the team configures itself, versioned alongside the deploy. Manage the agent like a team member: a target, an indicator and a performance review.
Docs, CLI and MCP.
One CLI for the whole cycle, from login to deploy, and native MCP support in both directions. Full documentation and real examples at docs.runflow.ai.
rf CLI: login, test locally with live reload, deploy staging → productionStaging and production isolated, with native Git.
Each agent lives in two isolated environments. Every change goes through the Studio flow, gets committed to Git and ships to production, with permissions by team. Whoever tests does not need deploy access.
One LLM portal for the whole company.
Centralized, governed access to every model. Smart routing, automatic fallback and cost control by team and by user, without each team signing up for its own API.
Service on the official WhatsApp, AI + human.
Agents serving customers on Meta's official WhatsApp, in the same thread as the human agent. The AI resolves what it can, and hands over to a person when the context demands it, without the customer noticing the switch.
Open architecture, built for enterprise.
No model lock-in, no cloud lock-in. The platform fits into your stack, not the other way around.
150+ connectors
ERPs, CRMs, APIs and channels. HubSpot, Twilio, Slack and your legacy systems, with MCP and internal tools.
SaaS or self-host
Managed by Runflow or in your own cloud (AWS, Azure, GCP), with your keys and your environments.
Knowledge (RAG)
Agentic RAG, a vector database and short and long term memory, connected to your data.
Security and governance at the standard your sector demands.
Isolated data
Per-client isolation, end to end. Runflow does not train models on your data, and you decide where it goes: providers with no retention, or your own model on your infrastructure.
PII sanitized
Sensitive data sanitized before it reaches the model. Brazilian data protection law (LGPD) and sector rules (SUSEP, ANS, professional secrecy).
Human approval
You approve before the agent acts wherever the risk demands it. Governance by design.
Audit trail
Every execution recorded and reproducible, ready for an audit.
Certifications in progress.
See the platform inside your operation.
Bring a real case. In a free conversation with an expert, we show where AI creates the most value and what the first agent in production would look like.