Perplexity wants a piece of the action. The search startup known for its citation-heavy answers now builds an internal AI coding assistant codenamed Teammate. It handles entire software projects from start to finish. And it aims straight at tools that already dominate developer workflows.
Details first emerged in a Business Insider report. The project sits in stealth. Engineers at the company test it daily. They use it to ship features faster than before. Yet the move signals something larger. Search alone no longer satisfies the ambitions of AI companies chasing revenue and relevance.
Cursor sits at the center of this fight. The AI-powered code editor passed $2 billion in annualized revenue earlier this year. Its parent company Anysphere commands valuations once reserved for much older software giants. Developers praise its speed. They value how it edits multiple files at once. Some call it the default choice for shipping code in 2026.
But competition heats up. OpenAI pushes Codex. Anthropic advances Claude Code. Microsoft refreshes GitHub Copilot with fresh models. Even Cognition Labs rebrands Windsurf into Devin Desktop. Each player promises agents that plan, write, test and debug without constant human input. Perplexity enters this fray with a different angle.
Its strength lies in research. The company built a name on accurate answers backed by sources. That same approach could translate to code. Imagine an agent that not only generates functions but also explains tradeoffs with citations from documentation and recent papers. Early internal tests hint at exactly that capability.
Recent posts on X show developers already mixing tools. One user outlined an AI team: ChatGPT for ideas, Perplexity for research, Claude Code for coding. Another noted the irony. Search companies now chase code. Chat interfaces chase code. IDE makers chase code. Why? Code offers the clearest proof that AI delivers real work.
Perplexity also invests in infrastructure. The company partners with NVIDIA on Vera CPUs. These chips power sandbox environments for its Perplexity Computer product. Early benchmarks show 1.5 times faster completion of coding tasks such as cloning repositories and running test suites. Concurrent sandboxes start nearly twice as fast. CEO Aravind Srinivas shared the progress on X just yesterday. The hardware edge may give Teammate an advantage in agentic workflows that require safe execution environments.
Market numbers tell a compelling story. Cursor’s rapid growth surprised many. It doubled revenue in months. Talks of a $50 billion valuation circulated in April. The editor evolved from simple autocomplete to full agentic systems. Background agents now run on isolated virtual machines. They test changes. They record videos and logs for review. A Bugbot add-on handles automated pull request checks for an extra fee.
Yet not every developer sticks with one tool. Recent rankings place Claude Code at the top for reasoning depth. Codex excels at raw generation. Cursor wins for everyday velocity. A January analysis from Faros.ai listed these three plus GitHub Copilot and Cline as front-runners. Choice depends on priorities: speed, control or autonomy.
Perplexity’s entry could shake this up. The company ditched its ad business to focus on enterprise. That decision raised eyebrows at a multi-billion valuation. It also freed resources. Enterprise buyers want reliability, provenance and integration with existing stacks. Perplexity’s citation habit addresses the first two. Sandbox improvements tackle the third.
So far the tool remains internal. No public launch date exists. No pricing details surfaced. But the codename Teammate suggests intent. This isn’t just another autocomplete plugin. It’s positioned as a colleague that owns projects end to end.
Developers on forums debate the shift. Threads from early 2026 already asked which AI coding product sees daily use. Many picked Claude Code for its agents. Others stayed loyal to Cursor for its polish. A few ran multiple in parallel. One Reddit user claimed Claude Code’s agents outperform even Cursor paired with top models.
Larger trends support Perplexity’s bet. Companies across tech now treat AI coding as table stakes. ServiceNow integrates with Codex, Claude Code and Cursor inside its platform. Replit raised $400 million at a $9 billion valuation on the back of its own Agent 4. Microsoft works on fresh models to regain ground after early stumbles with Copilot.
The pressure comes from both sides. Model makers want distribution. Tool makers want the best intelligence. Cursor CEO Michael Truell captured the dynamic in an interview. His company grabs top models from many providers, integrates them tightly, then focuses on user experience. Perplexity could follow a similar path. Its search index and real-time web access offer unique data advantages for code-related queries.
Challenges remain. Trust matters. Only 29 percent of developers fully trust AI accuracy according to surveys. Hallucinated code wastes time. Security reviews grow stricter. Enterprises limit agentic tools that touch production systems. Perplexity must prove its assistant avoids common pitfalls.
Its research-first DNA may help. By surfacing sources for every suggestion the tool could build confidence faster than pure generative rivals. Early internal adoption provides a proving ground. Engineers dogfood the product on real Perplexity codebases. Feedback loops tighten quickly.
And the timing feels deliberate. Cursor’s momentum continues. Valuation chatter refuses to die down. OpenAI eyes acquisitions in the space. Anthropic expands Claude Code’s reach. Perplexity refuses to watch from the sidelines. Instead it builds quietly. Then it prepares to enter with a product shaped by its own strengths.
Whether Teammate reaches the same adoption levels as Cursor remains unknown. The market grows fast enough for multiple winners. Yet the mere fact Perplexity pursues this direction validates the category. Code isn’t just another use case. It becomes the battlefield where AI companies prove their worth.
Watch for more signals soon. Infrastructure announcements around Vera CPUs already hint at deeper agent ambitions. Public demos could follow. When they arrive expect comparisons not just to search but to the full stack of coding agents now reshaping software development. The bar sits high. Perplexity arrives ready to compete.
WebProNews is an iEntry Publication