AI agents started 40% of all GitHub pull requests in September — 110 million of them. That number doubled in six months. Your review queue did not double in six months. This is the gap Tessl is now trying to close.
On November 6, Tessl launched its Software Factory Platform: a composable, agent-agnostic layer that sits above your coding agents and manages what they actually ship. Not another AI coding assistant. The factory floor.
Who Built This and Why It’s Worth Paying Attention To
Guy Podjarny built Snyk into an $8.5B company by solving a problem developers didn’t know they had: open-source dependency security. Tessl is the same bet, different layer. His thesis: developers have coding agents, but no system to govern what those agents produce. The company raised $125M from Index Ventures and Accel before shipping a single loop.
The core insight is deceptively simple. Podjarny puts it this way: “The challenge is getting agents to understand the boundaries of work, and when to stop. Verification is the fundamental accompaniment of increased agent autonomy.”
That’s not a marketing line. It’s the genuine problem. Agents will happily write code, open PRs, and declare victory. Whether the code is correct, secure, or consistent with your team’s standards is a separate question. Tessl’s answer is a verification layer baked into every workflow.
Three Loops, One Factory
Tessl’s platform is built around loops — repeatable, verifiable workflows combining agent skills with automated checks. The November 6 launch includes three ready-to-configure loops:
Ticket to PR
Assign a Linear issue to Tessl or add the tessl label to a GitHub issue. The agent reads the ticket, implements the change using your team’s versioned skill files as context, and opens a PR. Code Review and Vulnerability Remediation can run automatically before the PR ever hits your queue. Your team makes the final merge decision.
The trigger is designed to be trivially easy — one label, one assignment. The discipline is in the skill files that define what “done” looks like for your team.
Code Review
Every PR gets checked against your engineering standards automatically. Those standards live in versioned skill files in your own repo — not locked in Tessl’s platform. When your team updates a standard, every project using that skill inherits the update. When a better model ships, the same skills produce better reviews without anyone rewiring anything.
Vulnerability Remediation
When a scanner — Snyk, Semgrep, or Dependabot — flags a vulnerability in a PR, Tessl proposes a fix, then re-runs the scanner to verify the issue is gone. Not just unflagged — gone. Human review happens on the proposed fix rather than the original finding, which is significantly faster. Setup takes about seven minutes.
The Numbers (With a Caveat)
Tessl reports that 80% of PRs processed through their platform required no human editing, and cost per PR dropped from $75 to $13 — an 83% reduction. These figures come from Tessl’s own internal GTM workflows, not an independent case study. That caveat matters. But even a fraction of those gains would be significant for any team processing hundreds of agent-generated PRs per week.
The $75 → $13 figure covers the full cost of getting a PR from ticket to merge: agent compute, human review time, and rework. As Octoverse data shows, agent PR volume is on a steep upward curve. The difference between $75 and $13 compounds fast at scale.
No Lock-In, by Design
The platform’s credibility rests on one architectural choice: skills are stored as plain text files in your own repository. Tessl reads them; your agents read them; you own them. The platform works with Claude Code, Cursor, GitHub Copilot, Gemini, and others. If you decide Tessl isn’t the right fit, you leave with your entire skill library intact.
Most platforms prefer configuration to be opaque and exits to be painful. The fact that Tessl built it differently, backed by $125M in funding, suggests they’re betting on trust as a growth mechanism rather than lock-in. That’s worth something.
How to Get Started
New users get $200 in platform credits through November 30, 2026 (credits activate for 30 days after use). The documentation at docs.tessl.io walks through CLI setup and project initialization. The fastest path is enabling Vulnerability Remediation on one repository — seven minutes, immediate signal on whether your agent-authored PRs are introducing security issues you’re not catching.
If you’re already running coding agents in production, the question isn’t whether you need something like this. It’s which team in your org discovers the problem first.
