What is DeepSeek building to compete with Claude Code and Codex?

DeepSeek is building a native AI coding platform known internally as “Code Harness.” The product is designed to compete directly with coding agents such as Anthropic’s Claude Code and OpenAI’s Codex.

What is an agentic coding tool?

Agentic coding tools go beyond autocomplete by planning, writing, testing and debugging software with minimal human input. These systems aim to automate larger parts of the software development process rather than just suggesting snippets of code.

Why does DeepSeek want to build its own coding harness?

DeepSeek already provides AI models used inside coding tools, but owning the harness means controlling the full developer experience. That includes terminal interfaces, integrations, checkpoints, rollback systems and workflow automation where long-term customer loyalty is often built.

Why are AI coding assistants becoming such a major market?

AI coding tools are increasingly viewed as productivity infrastructure for software engineers and enterprises. Companies believe developers may eventually spend most of their workflow time interacting with AI agents rather than traditional coding environments.

How does DeepSeek’s pricing compare with competitors?

DeepSeek said its V4 Flash model costs about $0.14 per million input tokens, while Anthropic’s Claude Opus 4.7 reportedly costs around $15 per million tokens. That price gap could become important for businesses running large-scale autonomous coding systems.

What products is DeepSeek trying to compete against?

The company specifically referenced Claude Code, Codex, Cursor, GitHub Copilot, Manus and OpenClaw in recruitment listings. This suggests DeepSeek is targeting both enterprise AI coding platforms and developer-focused productivity tools.

Why does owning the “full stack” matter in AI?

Owning both the model and the interface allows companies to control user relationships, pricing and ecosystem integrations. Many AI firms believe the interface layer is where long-term monetisation and platform dominance will emerge.

What is DeepSeek V4 and why is it important?

DeepSeek V4 is the company’s latest large language model family released in April 2026. The models are designed for coding, reasoning and agentic workflows while competing aggressively on pricing against Western AI providers.

Image for illustrative purposes only. Not a real photo.

DeepSeek builds AI coding rival to Claude Code

Written by Liezl GambePublished May. 22 2026Last Updated May. 22 2026

DeepSeek is building a new AI coding platform designed to compete directly with Anthropic’s Claude Code and OpenAI’s Codex as competition intensifies in the agentic software development market.

The company began recruiting a Beijing-based “Code Harness” team responsible for developing a native coding agent platform that combines AI models with developer workflow tools such as terminal integrations, testing systems and rollback controls.

DeepSeek engineer Deli Chen said the internal strategy was based on the concept that “Model + Harness = Agent,” highlighting the company’s push to control both the underlying AI model and the interface developers use to interact with coding systems.

The recruitment listings specifically requested candidates with experience using products including Claude Code, Codex, Cursor, GitHub Copilot, Manus and OpenClaw, signalling that DeepSeek intends to compete across the broader AI coding ecosystem.

The expansion follows the release of DeepSeek V4 in April, with the company positioning its models as significantly lower-cost alternatives to leading Western AI systems for large-scale coding workloads and autonomous agent tasks.

DeepSeek said its V4 Flash model costs roughly $0.14 per million input tokens, compared with around $15 per million tokens for Anthropic’s Claude Opus 4.7, potentially giving the company a major pricing advantage for developers running continuous AI-assisted coding operations.

The development also reflects China’s broader push to build domestic AI infrastructure and software ecosystems as global competition intensifies around AI models, developer tooling and the future of autonomous software engineering platforms.