Refold AI (formerly Cobalt) is an AI-native integration platform where autonomous agents plan, build, test, and maintain enterprise integrations across a company's stack — SAP, Oracle, Workday, brittle APIs and 1,000+ connectors — replacing traditional middleware and manual glue code. Founded by Jugal Anchalia (CEO) and Abhishek Kumar (CPO), with Akshay J. as CTO, it is based in San Mateo and Bengaluru and raised a $6.5M seed in August 2025 co-led by Eniac Ventures and Tidal Ventures.
What was Cobalt, and why did it rebrand to Refold?
Cobalt began as an embedded iPaaS for shipping integrations faster. As the product evolved into an AI-native execution layer — autonomous agents that reason about, build, and maintain integrations rather than templates humans wire by hand — the company rebranded to Refold AI to reflect that shift. The rebrand marks the move from a connectors product to an execution layer agents act through.
How is Refold AI different from a legacy iPaaS like MuleSoft or Workato?
Legacy iPaaS tools ship templates and connectors that engineers still assemble and babysit by hand, while global systems integrators bill by the hour for custom edge cases. Refold turns every edge-case request into a repeatable, productized agent: Workflow Code Agents generate and maintain integration logic without boilerplate, MCP Chains let business teams describe an outcome in natural language, and self-healing workflows adapt when an underlying API changes — grounded in a context graph so the work compounds instead of breaking.
What is the context graph, and why does it matter?
The context graph is Refold's durable moat: a system that captures how an enterprise's systems behave and, crucially, why past integration decisions were made. A new model can rewrite the code but it cannot rewrite the reasons. By making that institutional memory queryable, Refold's agents act with the decision traces a real coworker accumulates over years — instead of guessing from a single prompt window — so integrations survive the people and code that produced them.
What problem does Refold AI solve, and who uses it?
By Refold's own framing, enterprises spend on the order of $350B a year keeping systems stitched together — the integration or 'duct-tape' tax (Refold's market framing, not an audited third-party statistic). Modern agents are blocked at that last mile because enterprise systems are heavily customized — think bespoke SAP — the data is messy, and the reasoning behind past decisions is lost. Refold serves solution-engineering teams, business teams, and SaaS product teams, replacing brittle middleware and manual glue code with autonomous integration agents grounded in a context graph. Early customers include Incorta.