Skip to content

Perplexity is quietly building an AI coding tool to take on Cursor and Claude Code

Chimera readability score 59 out of 100, Graduate reading level.

Perplexity could be the latest entrant to the red-hot AI coding wars.
The San Francisco-based AI search startup has built an internal AI coding tool that it may launch publicly later down the line, a person familiar with the matter said.
For now, Perplexity has codenamed the tool "Teammate," and its engineers have been using it since May, according to screenshots obtained by Business Insider.
It's unclear exactly if or when Perplexity, which was valued at $20 billion in a funding round last year, will launch the product. If it does, it would put Perplexity — originally focused on an AI-powered search engine that competes with Google — much closer to vying for supremacy with Cursor, Anthropic, and OpenAI, all of which have built widely-used AI coding products.
Teammate is meant to oversee software projects from start to finish, according to an internal announcement seen by Business Insider.
"It's built for long-horizon engineering work: owning projects, investigating issues, and monitoring services," the announcement reads.
A Perplexity spokesperson declined to comment.
Perplexity engineers have given the tool tasks such as finding bugs in internal systems, screenshots show.
The AI tool is model-agnostic, meaning that it's not built on any particular chatbot, the person familiar with the matter said.
Perplexity's chief technology officer, Denis Yarats, has also urged the startup's engineers to use AI for coding.
A few weeks before Teammate launched internally, the executive wrote in messages viewed by Business Insider that by the end of the year or sooner, software engineers should "stop looking at code" and just use AI.
Yarats also defended AI against accusations that it produces "slop," or poor-quality code.
"Slop is not going to be a thing" as long as the code it generates passes quality checks, Yarats wrote.
Have a tip? Contact Charles via email at crollet@businessinsider.com or on Signal and WhatsApp at 628-282-2811. Use a personal email address, a nonwork WiFi network, and a nonwork device; here's our guide to sharing information securely.

Sentinel — Human

Confidence

The text reads like standard business reporting, integrating insider knowledge with published evidence, exhibiting strong human journalistic provenance rather than pure AI generation.

Signals Detected
low severity: Moderate sentence length variance; tone is journalistic but leans toward direct reporting.
low severity: Flows logically from a specific development (Teammate) to broader competitive context and internal philosophy.
low severity: Attribution is primarily to an unnamed 'person familiar with the matter' and direct quotes/internal documentation, suggesting reliance on journalistic sourcing rather than pure data aggregation.
low severity: Claims rely heavily on citing specific internal sources (screenshots, internal announcements) which anchors the narrative in tangible evidence, making fabrication less likely than purely synthesized narratives.
Human Indicators
The inclusion of specific, verifiable references to 'screenshots obtained by Business Insider' and direct quotes attributed to an executive (Denis Yarats) suggest reliance on primary sourcing typical of investigative journalism.
The closing paragraph containing a security tip section is characteristic of online news outlets.