Palma.ai’s $1.8M bet on governing AI agents at runtime

Palma.ai’s September 2026 pre-seed round backs a specific enterprise AI problem: assistants can now take actions in company systems, but permissions and oversight can become fragmented when each assistant or integration is managed separately. Palma.ai proposes a shared layer between agents and those systems, rather than another assistant. 71131

Who invested, and where will the money go?

D11Z led the $1.8 million round, with Plug and Play Ventures, Deel and Scale Now Ventures participating. Angel investors included executives from Cisco and Deel; the reporting does not identify Cisco itself as an investor. Patrick Eden, CEO, and Julian Kolbe, CTO, co-founded Palma.ai. 7511

The company plans to expand its engineering and go-to-market teams as it targets larger organizations deploying agents across multiple platforms. Available reporting does not specify a budget split, hiring targets or a rollout timetable. 11

How does the shared governance layer work?

Palma.ai says its platform gives each authorized person or agent a governed connector carrying the tools they may use and Skills describing approved ways to perform work. It serves those capabilities through the Model Context Protocol (MCP), so MCP-capable clients can connect to the same gateway rather than requiring a separate integration for each assistant. Palma.ai lists Claude, Gemini and Copilot Studio among supported clients; compatibility in a particular enterprise still depends on its setup. 24262731

The distinction is between permission to reach a tool and permission to perform a particular action. Palma.ai says it evaluates a call’s actual arguments against policy, can hold higher-risk actions for human approval, and records calls—including denied or held requests—with information tying them to a person, agent and client. That approach is intended to let employees use approved capabilities while administrators retain centralized oversight. 242527

What remains to be verified?

Palma.ai describes its audit trail as tamper-evident and has published an account of hash-chained logging. Those are company claims, not independently verified guarantees in the available reporting. The sources also do not establish that every proposed workflow, such as CRM updates or financial reporting, is already deployed for customers. The funding supports Palma.ai’s plan to broaden adoption; it does not by itself demonstrate how the controls perform in a live enterprise environment. 243611