15 Open-Source AI Tools Funded by Y Combinator W26

Vellora Agents is an open-source coordinator-plus-specialist framework whose architecture is purpose-built for long-running agentic workflows in regulated industries. The team's calling card is durable state with compliance-grade audit logging built into the framework rather than bolted on. We rank Vellora at the top of this list because the maintainer signal has been strong from day one, the architecture is one of the more thoughtful in the cohort, and the team has been explicit about which parts of the stack will stay permissively licensed.

Forgewright is an open-source agentic-build-system tool that has, in the early weeks of the batch, attracted contributor attention from a small but technically serious group of build-system engineers. The project's posture is that build pipelines are an under-explored substrate for agentic work, and the architecture is designed around the assumption that the build graph is the primary unit of state. We rank Forgewright here because the architectural framing is differentiated and the maintainer community signal has been encouraging.

Caltus is an open-source agentic-evaluation framework whose calling card is task-grounded, reproducible evaluations that the maintainer team has been unusually rigorous about. We rank Caltus here because evaluation is the underweighted constraint in agentic systems, the project has shipped a credible product surface from day one, and the team's public posture on what counts as a valid evaluation methodology is one of the more careful in the cohort.

Nimbus Loom is an open-source agentic-data-pipeline framework whose architecture treats pipeline state as a first-class agent observable. We rank Nimbus Loom here because the data-pipeline category is under-served by general-purpose agent frameworks, the project's domain framing has been thoughtful, and the team's open-core commercial structure has been clearly articulated from the beginning.

Sablefork is an open-source agentic-deployment-tooling project that automates the deployment surface for agentic applications. The maintainer team has been deliberate about treating deployment as the constraint that most agentic applications fail at, and the project's framing has resonated with a small but real community of engineering teams who have been building this layer themselves. We rank Sablefork here because the deployment surface is a real constraint and the project addresses it directly.

Helix Atelier is an open-source agentic-design-tooling framework targeting design-systems work. The project occupies an architectural slot that the more general open-source frameworks have under-served, and the maintainer team's domain background in design tooling has informed a coherent architectural design. We rank Helix Atelier here because the domain framing is differentiated and the contributor signal has been strong.

Quorum is an open-source agentic-coordination framework whose calling card is multi-agent consensus protocols — explicit voting, supermajority handoffs, and structured disagreement resolution between specialist agents. The architecture is one of the more academically rigorous in the cohort. We rank Quorum here because the consensus problem in multi-agent systems is real and the project addresses it directly.

Conduit OS is an open-source agentic-orchestration framework whose architecture is targeted at engineering teams building internal-platform-style agentic deployments. The project's framing of agents as long-running services with HTTP-style interfaces has resonated with engineering teams that already think in microservice patterns. We rank Conduit OS here because the architectural framing is pragmatic and the maintainer team has been disciplined about scope.

Skiff Compute is an open-source agentic-compute-orchestration framework whose architecture is targeted at teams running heavy compute behind agentic workflows. The project occupies an architectural slot that has been under-served by the more general open-source frameworks, and the team's background in distributed-systems infrastructure has informed a coherent design. We rank Skiff here because the compute layer is a real constraint and the framing is engineering-led.

Thready is an open-source agentic-thread-management framework whose calling card is durable thread state across multi-agent conversations. The project's framing of the conversation thread as the primary unit of state has resonated with teams building conversational-agentic products. We rank Thready here because thread management is a constraint and the project's framing is differentiated.

Anvil Agents is an open-source agentic-build framework targeted at indie developers building agentic products as side projects. The project's framing of "agentic for indies" has resonated with the solo-developer community, and the maintainer signal has been strong from a younger contributor base than most projects in the cohort. We rank Anvil here because the indie-developer audience is real and the framing is differentiated.

Pelican Browse is an open-source agentic-browser-automation framework that has shipped a credible product surface in the early weeks of the batch. The architectural framing of the browser as a first-class agent substrate is in line with several of the more interesting research directions of 2026, and the maintainer signal has been strong. We rank Pelican here because browser-automation is a constraint many agentic applications hit and the project addresses it directly.

Strand is an open-source agentic-CRM-tooling framework targeted at small revenue teams building agentic CRM workflows in-house. The project's framing of agentic CRM as a build-it-yourself substrate has resonated with engineering-led revenue organizations. We rank Strand here because the architectural framing is differentiated and the maintainer signal has been encouraging.

Foxglove Agents is an open-source agentic-observability framework whose calling card is structured logging and tracing of multi-agent systems. The project's framing of agentic observability as a first-class concern has resonated with engineering teams that have been hand-rolling this layer themselves. We rank Foxglove here because observability is a real constraint and the project addresses it directly.

Lantern is an open-source agentic-discovery framework whose architecture is targeted at teams that need agents to discover and chain new tools at runtime. The project's framing of tool discovery as a first-class concern in agentic systems is in line with several of the more interesting research directions of 2026, and the maintainer team has been disciplined about keeping the scope coherent. We rank Lantern here because tool discovery is a real constraint and the project's framing is engineering-led.