Ramp opens Inspect, its collaborative AI coding agent, to early access

Ramp is opening early access to Inspect, a cloud-based coding agent that can take tasks from Slack, GitHub, Linear, the web or mobile and prepare a pull request for review. The notable twist is that Ramp says Inspect already accounts for 74% of its merged pull requests, though that is the company’s own usage figure, not an independent measure of code quality.

Ramp Watch analysis

What happened

Inspect works in a team’s development environment and can turn a written task into a visually validated, reviewed pull request, according to Ramp. Sessions are collaborative by default, so colleagues can add context or steer the work before review. Ramp says users can choose a model through Ramp Router and turn useful ways of working into agents or automations.

The company says 99% of its engineering, design and product teams used Inspect in the past month. It also says its builders now ship 3.4 times as many pull requests per person as a year ago, with Inspect a major reason. Ramp announced early access on 27 October in its Inspect announcement.

Why it matters

Inspect’s pitch is not simply “AI writes code”. It is that a coding agent can meet teams where work already happens, use their existing development setup and leave a pull request for people to review. That could make software changes easier to start across a company, not just for developers already comfortable running an agent locally.

Ramp’s adoption and productivity figures are striking, but they are company-reported and do not tell us whether the resulting code is better, safer or cheaper to maintain. A large share of merged pull requests is evidence of use, not a quality score. The human review step is doing important work in this story.

Our read

The most interesting test is whether Inspect helps more people turn useful ideas into well-reviewed changes without creating a new queue of code for engineers to clean up. Ramp’s internal adoption suggests the workflow has caught on there; teams outside Ramp will want to see what it takes to make that happen in their own codebases.

What to watch

• Who gets access during the early-access rollout, and when.
• Whether teams can use their own models and development environments as described.
• Independent evidence on code quality, review burden and maintenance costs.
• Whether Ramp publishes results beyond adoption and pull-request volume.

Discussion spark: If an AI coding agent produces most of a company’s merged pull requests, should success be measured by the volume shipped or by the quality and maintenance cost of the code?

not affiliated with or endorsed by Ramp