Turn prompts into pipelines
The open-source orchestration layer for LLM workflows. Build agents from the ground up. Run where data lives.
Why AgentFlow?
Most agent platforms lock you in and send your data to external servers. AgentFlow is built for complete control, absolute security, and zero vendor lock-in.
Security-First
Run where your data lives. Your pipeline runs inside your secure cloud or on-premise infrastructure. No leakage, no external processing.
No Vendor Lock
Our stack is completely open-source. Build pipelines that belong to you, without recurring fees or dependencies on proprietary platforms.
Your Stack
Bring your own infrastructure, custom tools, and models. Connect local LLMs, commercial APIs, or open models without friction.
How It Works
Go from raw prompts to robust autonomous LLM workflows in three simple steps.
Connect
Link your local models, vector stores, API keys, and custom developer tools.
Configure
Define routing rules, agent profiles, session states, and tool bindings in clean YAML.
Deploy
Spin up your agent workspace, connect it to your app, or run it headless as a CLI daemon.
Core Capabilities
AgentFlow comes pre-loaded with powerful tools, memory modules, and multi-agent routing engines.
Multi-Provider Routing
Seamlessly route tasks to OpenRouter, Gemini, Codex, or local instances via LM Studio. Choose cost, speed, or intelligence dynamically.
Specialized Agent Profiles
Load pre-configured profiles like TARS, research assistants, or coder agents. Define personality, temperature, system constraints in YAML.
Protocol Tool Integrations
Equipped with full Model Context Protocol (MCP) support. Bind external databases, code runtimes, or custom files as execution tools.
Skills & Semantic Memory
Session-aware vector caches preserve long-term context and let agents learn skills dynamically. Maintain context across multiple conversation boundaries.
Telegram Bot Integration
Run your workflows directly from Telegram. Chat with agents, trigger scheduled execution flows, and receive notifications via inline commands.
Local-First Isolation
Keep code execution, configuration registries, and state storage completely local. Perfect for offline tasks and sensitive data projects.
Flexible Pricing Plan
Scale your agent operations with transparent, affordable plans. All options include the open-source CLI engine.
Free
- Basic MD prompt builder
- Git workflow sync
- GitHub open-source stack
- Local-only configuration
Premium
- Multi-AI model routing
- Cloud storage & sync
- Saved template library
- Structured JSON outputs
- Advanced skills support
Pro
- Preloaded orchestration runtimes
- Priority support channels
- Custom skills authoring
- Team-wide workspaces