Tessl launches AI software factory for engineering teams
Tue, 17th Nov 2026 (Today)
Tessl has launched a software factory platform for AI-assisted development, aimed at engineering teams that want to hand more work to software agents.
The London startup says the platform gives developers a structured way to turn code repositories into repeatable workflows for tasks such as implementing tickets, reviewing code and fixing vulnerabilities. It is built around reusable "skills" that define how agents should carry out a task, and "loops" that combine those skills into workflows with verification steps.
The launch comes as software teams test broader use of AI coding tools while remaining cautious about trusting their output. Tessl argues that the constraint is shifting from generating code to checking whether agent-produced code is reliable, secure and worth the cost.
Founded in 2024 by Snyk founder Guy Podjarny, Tessl has raised USD $125 million from investors including Index Ventures, Accel, GV and boldstart. It describes its system as agent-agnostic and model-agnostic, allowing teams to build workflows without tying themselves to a single AI model or coding agent.
Workflow focus
Instead of relying on ad hoc scripts to coordinate AI tools, the platform uses a declarative approach to define software development workflows. Teams can add checks and feedback mechanisms so agent output is reviewed against repository rules and team standards.
Its initial templates cover three main use cases: Ticket to PR, Code Review and Vulnerability Remediation. In the first, an agent implements a scoped issue and opens a pull request. In the second, changes are checked against engineering standards. In the third, scanners including Snyk identify vulnerabilities introduced by a pull request, after which Tessl proposes or applies fixes and then rescans the code.
Tessl says it has already used the platform internally. It says 80% of one week's pull requests handled by the system required no human hand-editing, while the cost per pull request fell from USD $75 to USD $13 across implementation, review, security fixes and related workflow steps.
Those figures highlight the commercial question facing companies expanding their use of AI in software engineering. Coding assistants can produce drafts quickly, but businesses still need to manage review workloads, security exposure and cloud spending tied to repeated model calls.
Trust question
The debate in software development has moved beyond whether AI can generate code at all. Attention is increasingly turning to governance: how to define the boundary of a task, stop agents from overreaching, and verify that output meets technical and security requirements before it reaches production systems.
That is the problem Tessl is trying to address with a factory model rather than a single assistant working with an individual programmer. The concept shifts the focus from one-off prompts to repeatable processes, with humans setting rules and checking outcomes while agents carry out more of the routine work.
Guy Podjarny, founder of Tessl, set out that view in comments accompanying the launch. "Engineering and developer teams aren't finding it difficult to assign work to agents. The challenge is getting agents to understand the boundaries of work, and when to stop," he said.
He added: "Verification is the fundamental accompaniment of increased agent autonomy. Tessl Factory strikes the right balance to get agent delegation off the ground with an appropriately gradual approach set by the pace of teams running on Tessl."
Tessl is also offering USD $200 in usage credits to teams trying the platform for the first time. The incentive suggests it is seeking early adoption among development teams interested in agent-based workflows but unwilling to commit to broader deployment without first testing the economics and review burden.
Podjarny is best known for founding developer security group Snyk and later serving as chief technology officer at Akamai after the acquisition of his earlier startup Blaze. His return with Tessl reflects growing investor and industry interest in tools that sit above individual coding assistants and aim to orchestrate larger parts of the software development lifecycle.
For software teams, the appeal of that model is straightforward: if more ticket handling, code review and remediation can be pushed to agents under clear controls, engineers may spend less time on repetitive work. The open question is whether those controls can scale as organisations move from experimenting with AI-generated code to relying on it in day-to-day delivery.