Frequently asked questions

No. The managed hosting at app.openlegion.ai requires no coding — start a free trial, pick a template, add your LLM API key, and your agents are live in minutes. The self-hosted version requires Python 3.10+ and Docker. Either way, the built-in team templates (Dev Team, Sales Pipeline, Content Studio) work out of the box with no configuration needed.

Yes — you bring your own API keys from Anthropic, OpenAI, Google, or any of 100+ supported providers. You pay providers directly at their published rates. OpenLegion charges for the platform and infrastructure only. There is zero markup on model usage.

Any task a human performs on a computer with a browser or terminal. Agents can browse and interact with any website, log into web applications, fill out forms, extract data from any page, send emails and messages, manage files and folders, write and execute code, process documents, post to social platforms, monitor pages for changes, and run custom automations — all 24/7 without supervision. Common deployments include sales outreach pipelines, competitive research, lead qualification, developer workflows, invoice processing, content production, and internal task automation.

OpenLegion is a production-grade AI agent framework and platform that deploys autonomous agent fleets in isolated Docker containers. Each agent gets its own budget, permissions, and secrets vault — with six security layers enabled by default. It requires only Python, Docker, and an API key. No Redis, no Kubernetes, no LangChain. Licensed under BSL 1.1 (source-available).

OpenLegion container-isolates every agent, proxies all credentials through a vault, and enforces per-agent budgets — most frameworks don't. CrewAI and similar frameworks run agents in shared processes with no isolation, no cost controls, and API keys stored in config files. OpenLegion uses deterministic YAML workflows instead of letting an LLM decide task routing, making execution predictable and auditable.

100+ LLM providers through LiteLLM. This includes Anthropic (Claude), OpenAI (GPT), Google (Gemini), Mistral, Moonshot, and any OpenAI-compatible API. You can assign different models to different agents in the same fleet — no vendor lock-in.

Through blind credential injection — agents never see API keys. Keys are stored in a vault on the mesh host. When an agent calls an LLM, the request goes through a vault proxy that injects the credential at the network layer, tracks token usage, and enforces budget limits. Even a fully compromised agent cannot access your API keys.

No. OpenLegion runs on a single machine with no external services. You need only Python 3.10+, Docker, and an LLM API key — no Redis, no Kubernetes, no LangChain, no external databases.

Yes — OpenLegion is designed for enterprise deployment. It includes on-premises support (including air-gapped environments), deterministic YAML workflows, per-agent cost governance, role-based access controls, credential isolation via vault proxy, and an audit-ready codebase of ~30,000 lines with 2,100+ tests. All security layers are enabled by default.

Yes. The full stack runs entirely on your own infrastructure with no external dependencies beyond Docker and Python. No data leaves your network. The coordinator, agents, vault, and dashboard all run on a single machine, making it suitable for air-gapped, regulated, and on-premises environments.

An AI agent platform is managed infrastructure for deploying, orchestrating, and governing autonomous AI agents in production. Unlike raw framework libraries that only provide agent logic primitives, a platform handles container isolation, credential vaulting, per-agent cost controls, observability, and deployment — so teams can ship agents without building DevOps from scratch.

A framework is a code library for building agent logic — tools, prompts, memory, and chains. A platform adds operational infrastructure on top: container isolation, credential vaults, per-agent budgets, deployment pipelines, and fleet-wide observability. OpenLegion is both: a Python framework for authoring agents and a self-hosted platform for running them in production with security and cost governance built in.

OpenLegion uses deterministic YAML DAG workflows — no LLM sits in the control plane deciding who does what. You define task graphs with sequential and parallel execution patterns, coordinated through a centralized Blackboard and pub/sub messaging. The orchestrator routes tasks based on the DAG definition, making every execution path predictable, repeatable, and auditable. Fleet model, not hierarchy — agents execute their assigned steps and report results back to the coordinator.

AI agent security addresses the unique threats autonomous agents introduce: credential leakage, prompt injection, resource abuse, and data exfiltration. OpenLegion's six-layer defense — runtime isolation, container hardening, credential separation, permission enforcement, input validation, and Unicode sanitization — mitigates each threat independently, so a breach in one layer does not compromise the others.