Saoud Rizwan is the founder and chief executive of Cline, an open-source coding agent that can inspect repositories, edit files, execute commands, and use external tools under developer supervision. His defining conviction is that auditable coding agents should give programmers control over their models, infrastructure, costs, and the actions performed on their behalf.
Earlier in his career, Rizwan worked as an iOS developer at Yummygum and created Disk, an open-source Swift library for storing application data and images. His other Swift projects included CardSlider, DynamicJSON, Dance, and Piano.
In July 2024, he built Cline’s original prototype, Claude-Dev, for an Anthropic hackathon while working from his mother’s garage in Terre Haute. The project did not win, but its open-source release attracted developers interested in agents capable of operating across an entire codebase. Community feedback shaped features including custom rules and planning workflows, and some early contributors joined the company.
By July 2025, Cline had raised $32 million in seed and Series A funding, led by Emergence Capital and Pace Capital. In January 2026, Rizwan announced five million installations and $1 million in open-source grants. Cline expanded from its original VS Code extension into JetBrains, command-line, and software-development-kit workflows.
- Plan-and-Act workflows: After observing developers ask agents to outline changes before writing code, Rizwan separated planning from execution. The agent first explores the repository, gathers context, and asks questions; implementation begins when the developer approves the direction.
- Focus Chain: Rizwan argues that larger context windows do not prevent agents from losing track of requirements. His continuously updated task-planning approach keeps completed and outstanding steps visible throughout execution, favoring relevant context and explicit plans over indiscriminate prompt stuffing.
- Real-world agent evaluation: He judges models by whether they solve actual repository problems, remove dead code, pass builds, and verify their work. In one Cline comparison, GLM used more tokens than Opus but cost less and performed additional cleanup and build verification. His response to reported Databricks evaluations likewise prioritizes proprietary-codebase results and cost over public benchmark rankings.
- Open-weight models and inference choice: Rizwan considers model portability essential protection against vendor lock-in. His work adapting Cline to GLM-4.6 identified meaningful differences among providers serving the same model. He expects competition, caching, batching, and specialized infrastructure to reduce inference costs, and has urged American AI laboratories to release more open-weight models.
- Software supply-chain security: Rizwan emphasizes visible tool calls, reviewable commands, human approval, and provenance-backed publishing. After an unauthorized Cline CLI release in February 2026, his incident postmortem traced the breach through prompt injection, cache poisoning, and incomplete credential rotation.
Rizwan regards automated low-quality contributions and maintainer overload as genuine threats to open-source collaboration. He argues that inspectable code, portable models, verification systems, and competitive inference infrastructure become more valuable as coding agents gain autonomy.
