Open-source infrastructure for embedding AI agents in your product: sessions, sandboxes, tools and credentials included.
Homepage hero
1 / 3Overview
What is Opengeni?
Agents in your product. Infrastructure out of the box.
Opengeni is open-source AI infrastructure for putting agents inside your product — streaming, durable sessions, sandboxes, tools, credentials and React components, so you don't have to rebuild them. It was extracted from two years of running AI agents against real production cloud infrastructure.
Sessions survive errors with replayable history, agents execute in sandboxes with scoped short-lived credentials, any API or MCP server can be wired in as a tool, and you can use OpenAI, Azure OpenAI, OpenRouter or any OpenAI-compatible endpoint with no lock-in. Built-in dashboards trace every step, token and dollar.
The software is Apache-2.0 and free to self-host (Helm chart for any Kubernetes, Terraform for AWS/Azure/GCP). The managed cloud costs model spend + 5% with no platform fee, seat fees or minimums, and launch promos have included $100 in cloud credit.
Platforms and languages
Opengeni Availability
Platforms
Languages
Capabilities
Opengeni Key Features
Durable sessions
Sessions survive disconnects and resume from event points with replayable history.
Sandboxes
Agent code runs in isolated environments, separate from user secrets.
Credentials
Per-user OAuth tokens handled automatically with scoped, short-lived tokens.
Tools and MCP
Wire any API (OpenAPI YAML) or MCP server in as agent-callable tools.
Any model
OpenAI, Azure OpenAI, OpenRouter or any OpenAI-compatible endpoint — no lock-in.
Best for
Who uses Opengeni?
Developers and product teams
Embed AI agents in their own product without rebuilding sessions, sandboxes and auth
Plans and access
Opengeni Pricing
Freemium
Open source $0 license, self-host free; Cloud is model cost + 5%, no seat or platform fees.
Common questions
Opengeni FAQs
Is self-hosting really free?
Yes — the software is Apache-2.0 with a $0 license; you pay only your own infrastructure and model provider costs.
Is there per-seat pricing on the cloud?
No — the cloud costs model spend + 5% only, with no seat fees, platform fee or minimum commitment.
Can I run it on my own infrastructure?
Yes — a Helm chart runs on any Kubernetes, with Terraform for AWS, Azure and GCP.
Where are the technical requirements documented?
The docs cover local setup, production self-hosting, authentication, compute targets and product integration.
Reviews
See what the community thinks and share your experience.
—
Based on 0 ratings
Rating distribution
Community reviews
No reviews
No written reviews yet.
