Amp is Sourcegraph's take on agentic coding: no model picker, no knobs — it always runs frontier models with maximum reasoning and leans into autonomy. Work happens in shareable threads across the VS Code extension and CLI, with subagents for parallelizable work and team visibility into how colleagues prompt. It's deliberately opinionated and token-hungry; credits are consumed at cost, so sustained heavy use gets expensive, and there's no BYO-key escape hatch.
Bars show editorial scores from hands-on testing. Percentages showAmp's live community win rate in head-to-head votes.
Every agent on this page runs on Standard Compute