Greptile uses GPT-6 Astra to test user workflows and secure code
Greptile, an AI agent company, is leveraging OpenAI's GPT-6 Astra to expand its bug detection capabilities beyond static code review. Co-founder and CEO Daksh Gupta explains that the model's computer use features allow Greptile to automate application testing at scale. By starting development servers and simulating user interactions, such as resizing browsers to mobile viewports, the system can…
Key points
- Greptile uses GPT-6 Astra's computer use to simulate user workflows and spot visual bugs.
- Daybreak is used to provide startups with sophisticated security defenses against AI-powered attacks.
- Thousands of customers connect Codex to Greptile's Knowledge Base via MCP to reduce code reverts.
The company also utilizes OpenAI's Daybreak model to strengthen security defenses, aiming to make sophisticated protection accessible even to startups. Gupta notes that more powerful models increase attack risks, so Greptile seeks to ensure customers have defense mechanisms that match or exceed adversarial capabilities. Additionally, Greptile maintains a Knowledge Base that indexes codebases and documents them for both humans and agents. This documentation is shared with OpenAI's Codex via a Model Context Protocol (MCP) server, enabling Codex to write code that aligns with existing customer patterns. Gupta states that thousands of customers have connected Codex to this knowledge base, resulting in fewer reverts of AI-generated code and significant time savings.
Model page: GPT-6 Astra →
Greptile takes bug detection beyond code review with GPT-6 Astra
OpenAI · 20 October 2026
Loading the full article…
This text was published by OpenAI. It is reproduced here with attribution so you can read it in full; the rights remain with the publisher. Read it at the source ↗
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Agents & Tools
All →- Claude Haiku 5.5: New Features, Pricing and Efficient AI Workflows · 2 src
- OpenAI agents’ leaders say interfaces will remain despite automation progress · 1 src
- Entire launches Git forge for agent-human collaboration with Trail scoring · 1 src
- Amazon Bedrock adds prompt caching to reduce AI agent token costs · 1 src
- Nace AI releases Drex for Search with 92.05% on BrowseComp-Plus · 1 src
Comments
via GitHub Discussions