Agentic artificial intelligence infrastructure startup Modiqo Inc. says it can assist companies in moving from experimental workflows to repeatable and production-ready systems after raising $3 million in pre-seed funding.

Heavybit and Seligman Ventures co-led the round, which also saw participation from Irregular Expressions and a number of angel investors.

The startup is trying to fix the unpredictable nature of AI agents, which tend to be very creative and do some very strange and unique things that don’t always result in the outcome users had hoped for. What’s more, they do this even if they’ve successfully executed the same task the day before.

“Today’s AI agents are impressive in the demo and unreliable in production,” said Heavybit General Partner Joseph Ruscio.

According to Ruscio, the problem with AI agents is that every time they attempt to perform a task, they’re working from a one-off chat log rather than an artifact that can be proven and iterated on. In other words, each time they’re told to do something, they start again from scratch, even if they’ve done it before, and they may not make the same decisions, partly because of the constant changes to the models that power them and the systems they run on.

“Teams pay the same rediscovery tax over and over, and the abstractions emerging around them weren’t built to survive change,” Ruscio explained.

Modiqo’s solution to this is a local execution layer it calls Rote, which is designed to capture what happens during AI agent runs, so that successfully executed tasks can essentially be remembered and transformed into reusable workflows.

“Modiqo’s Rote captures what worked the first time and turns it into deterministic code — cutting token usage, reserving inference for the parts that genuinely need it, and reining in the context bloat plaguing today’s agent stacks,” Ruscio added.

Rote is aimed at systems operators who are tasked with ensuring that AI agent deployments run smoothly. The platform can connect AI agents to any kind of application programming interface, with no need for any server infrastructure, sidecars or software development kits. Once connected, it can prevent teams from having to repeatedly debug the same failures each time a model or API is updated, the company explained. Users will be able to see exactly what each agent executed, when it did it, why it acted the way it did and how much that workflow cost.

Founder and Chief Executive Chetan Conikee (pictured) said Rote has proved itself useful for early adopters because of the incredibly fragile nature of modern AI workflows. While it might seem easy to get an AI agent to perform a task the first time, users soon find that getting it to repeat that task if there’s any slight variation is next to impossible.

“Keeping it working the same as models, APIs and systems evolve is definitely not easy,” he explained. “Teams stitch together prompts, skills, scripts and integrations that quietly break as context is lost and systems change, forcing them to rebuild the same workflows again and again. Rote turns successful executions into durable, reusable assets.”

Photo: Modiqo