Qodo 2.1 solves your coding agents' 'amnesia' problem, giving them an 11% precision boost

As AI-powered coding tools flood the market, a critical weakness has emerged: by default, as with most LLM chat sessions, they are temporary — as soon as you close a session and start a new one, the tool forgets everything you were just working on.

Developers have worked around this by having coding tools and agents save their state to markdown and text files, but this solution is hacky at best.

Qodo, the AI code review startup, believes it has a solution with the launch of what it calls the industry's first intelligent Rules System for AI governance — a framework that gives AI code reviewers persistent, organizational memory.

The new system, announced today as part of Qodo 2.1, replaces static, manually maintained rule files with an intelligent governance layer. It automatically generates rules from actual code patterns and past review decisions, continuously maintains rule health, enforces standards in every code review, and measures real-world impact.

For Itamar Friedman, CEO and co-founder of Qodo, the release represents a pivotal moment not just for his company but for the entire AI development tools space.

"I strongly believe that this announcement of ours is most important we ever done," Friedman said in an interview with VentureBeat.

The 'Memento' problem

To explain the limitation of current AI coding tools, Friedman invokes the 2000 Christopher Nolan film Memento, in which the protagonist suffers from short-term memory loss and must tattoo notes on his body to remember crucial information.

"Every time you call them, it's a machine that wakes up from scratch," Friedman said of today's AI coding assistants. "So all it can do is, before it goes to sleep and restart, it could write whatever it did in a file."

This approach—saving context to markdown files like agents.md or napkin.md—has become a common workaround among developers using tools like Claude Code and Cursor. But Friedman argues this method breaks down at enterprise scale.

"Think about heavy duty software where you now have, let's say, 100,000 of those sticky notes," he said. "Some of them are sticky notes. Some of them are huge explanations. Some of them are stories. You wake up and you get a task. The first thing that [the AI] is doing is statistically starting to look for the right memos... It's much better than not having it. But it's very random."

From stateless to stateful

The evolution of AI development tools has followed a clear trajectory, according to Friedman: from autocomplete (GitHub Copilot) to question-and-answer (ChatGPT) to agentic coding within the IDE (Cursor) to agentic capabilities everywhere (Claude Code). But he contends all of these remain fundamentally stateless.

"In order for software development to really revolutionize how we do software development for real world software, it needs to be a stateful machine," Friedman said.

The core challenge, he explained, is that code quality is inherently subjective. Different organizations have different standards, and even teams within the same enterprise may approach problems differently.

"In order to really reach high level of automation, you need to be able to customize for the specific requirements of the enterprise," Friedman said. "You need to be able to provide code in high quality. But quality is subjective."

Qodo's answer is what Friedman describes as "memory that is built over a long time and is accessible to the coding agents, and then they can poke and check and verify that what they're actually doing is according to the subjective needs of the enterprise."

How Qodo's Rules System works

Qodo's Rules System establishes what the company calls a unified source of truth for organizational coding standards. The system includes several key components:

• Automatic Rule Discovery: A Rules Discovery Agent generates standards from codebases and pull request feedback, eliminating manual authoring of rule files.

• Intelligent Maintenance: A Rules Expert Agent continuously identifies conflicts, duplicates, and outdated standards to prevent what the company calls "rule decay."

• Scalable Enforcement: Rules are automatically enforced during pull request code review, with recommended fixes provided to developers.

• Real-World Analytics: Organizations can track adoption rates, violation trends, and improvement metrics to prove standards are being followed.

Friedman emphasized that this represents a fundamental shift in how AI code review tools operate. "It's the first time that AI code review tool is moving from reactive to proactive," he said.

The system surfaces rules based on code patterns, best practices, and its own library, then presents them to technical leads for approval. Once accepted, organizations receive statistics on rule adoption and violations across their entire codebase.

A tighter connection between memory and agents

What distinguishes Qodo's approach, according to Friedman, is how tightly the rules system integrates with the AI agents themselves—as opposed to treating memory as an external resource the AI must search through.

"At Qodo, this memory and agents are much more connected, like we have in our brain," Friedman said. "There's much more structure to it... where different parts are well connected and not separated."

Friedman noted that Qodo applies fine-tuning and reinforcement learning techniques to this integrated system, which he credits for the company achieving an 11% improvement in precision and recall over other platforms, successfully identifying 580 defects across 100 real-world production PRs.

Friedman offered a prediction for the industry: "When you look one year ahead, it will be very clear that when we started 2026, we were in stateless machines that are trying to hack how they interact with memory. And we will have a very coupled way by the end of 2026, and Qodo 2.1 is the first blueprint of how to do that."

Enterprise deployment and pricing

Qodo positions itself as an enterprise-first company, offering multiple deployment options. Organizations can deploy the system entirely within their own infrastructure via cloud premise or VPN, use a single-tenant SaaS option where Qodo hosts an isolated instance, or opt for traditional self-serve SaaS.

The rules and memory files can reside wherever the enterprise requires—on their own cloud infrastructure or hosted by Qodo—addressing data governance concerns that enterprise customers typically raise.

On pricing, Qodo is maintaining its existing seat-based model with usage quotas. At present, the company offers three pricing tiers: a free Developer plan for individuals with 30 PR reviews per month, a Teams plan at $38 per user per month (with 21% savings available for annual billing) that includes 20 PRs per user monthly and 2,500 IDE/CLI credits, and a custom-priced Enterprise plan with contact-us pricing that adds features like multi-repo context awareness, on-prem deployment options, SSO, and priority support.

Friedman acknowledged the ongoing industry debate about whether seat-based pricing makes sense in an age of AI agents but said the company plans to address this topic more comprehensively later this year.

"If you get more value, you pay more," Friedman said. "If you don't, then we're all good."

Early customer response

Ofer Morag Brin of HR technology company Hibob, an early user of the Rules System, reported positive results in a press statement Qodo shared with VentureBeat ahead of the launch.

"Qodo's Rules System didn't just surface the standards we had scattered across different places; it operationalized them," Brin said. "The system continuously reinforces how our teams actually review and write code, and we are seeing stronger consistency, faster onboarding, and measurable improvements in review quality across teams."

Founded in 2018, Qodo has raised $50 million from investors including TLV Partners, Vine Ventures, Susa Ventures, and Square Peg, with angel investors from OpenAI, Shopify, and Snyk.