- Letta — open-source agent framework and server (born from MemGPT) for building stateful AI agents.
Why fits: founded 2024; YC S24; monetizes via Letta Cloud for managed agent infrastructure.
- R2R (SciPhi) — open-source RAG engine and API for building AI apps with semantic search, ingestion, and retrieval.
Why fits: founded 2024; YC S24; sells SciPhi Cloud as the managed version.
- Composio — open-source integration platform that gives AI agents managed tools, actions, and auth connections.
Why fits: founded 2024; YC S24; cloud monetization through hosted agent integrations.
- Julep — open-source platform for building, deploying, and managing AI agents.
Why fits: founded 2024; YC S24; offers a managed Julep Cloud.
- Laminar — open-source LLM observability, tracing, and evals platform.
Why fits: founded 2024; YC S24; monetizes via hosted cloud observability.
- Stack Auth — open-source authentication and user management infrastructure.
Why fits: founded 2024; YC S24; hosted cloud version for teams that don’t want to self-host.
- Arcjet — open-source security-as-code SDK covering rate limiting, bot protection, and WAF-style controls.
Why fits: founded 2024; YC S24; cloud monetization for managed security policies.
- Hatchet — open-source task queue and background job orchestration for durable execution.
Why fits: founded 2024; YC S24; offers Hatchet Cloud as the managed control plane.
- Restack — open-source orchestration platform for AI agent workflows and distributed services.
Why fits: founded 2024; YC S24; monetizes through Restack Cloud.
- Triplit — open-source real-time database with client/server sync and offline support.
Why fits: founded 2024; YC S24; Triplit Cloud is the managed sync/database offering.
White-space gaps I noticed while sourcing:
- Open-source usage-based billing / cost metering for LLM and AI-agent workloads. Most tools in this area are older or proprietary, yet this is becoming critical infrastructure as AI apps scale.
- Open-source policy/guardrails-as-code for AI agent actions and tool calls. There are new agent frameworks and security SDKs, but almost no newer open-source layer specifically enforcing fine-grained agent permissions, tool-level auth, and compliance policies.