Amp

Amp is Sourcegraph's agentic coding tool, launched in 2025. The strategic context matters: Sourcegraph has spent nearly a decade building Code Search — an indexing engine that handles 10M+ LOC codebases with semantic search across files, repos, and language ecosystems. Amp is the agentic layer on top of that infrastructure. When the agent needs context about a function, a class, or a usage pattern, it queries Sourcegraph's pre-built index rather than scanning files at runtime. For teams already running Sourcegraph, this means faster and more accurate context retrieval than tools that have to discover code structure on every prompt.

The differentiator is real for one specific shape of team: enterprises with large codebases (often 1M+ lines, multiple services, deep dependency graphs) that already pay for Sourcegraph. Amp inherits the context layer those teams have already invested in. For solo developers and small projects, that advantage doesn't compound — the index isn't there yet, and the standalone value vs Cursor or Claude Code is less obvious. The pricing reflects this positioning: $19–$49/month for individuals on the paid tier (with a meaningful free tier for evaluation), with enterprise team pricing bundled with Sourcegraph Code Search.

Real cost math for the different user types. Individual evaluator using Amp's free tier: $0/month, generous enough to evaluate whether the Sourcegraph context layer is worth the team-tier upgrade. Solo paid user on $19/month: similar to Cursor Pro, with the Sourcegraph context advantage but less polished IDE-first experience. Enterprise team: pricing varies based on Sourcegraph contract size, typically $50–$100/user/month bundled with Code Search. The unit economics genuinely favor Amp at the enterprise tier; at the solo tier, Cursor or Cline usually wins on price-per-feature.

Amp vs Cursor — the head-to-head most builders evaluate. Cursor has the polished IDE-first experience, multi-model selection (Claude, GPT, Gemini), and the largest paying user base. Amp has the Sourcegraph context-graph advantage for large codebases and the enterprise compliance story (SOC 2, on-prem deployment) that Cursor doesn't ship out of the box. For solo work and small-to-mid codebases, Cursor wins. For enterprise teams at scale where context retrieval on 1M+ LOC matters, Amp is genuinely better. The dividing line is roughly: if your codebase fits in Claude's 1M-token context window, you don't need Amp's advantage.

Amp vs Augment Code — the other large-codebase specialist. Both target enterprise teams where context-window limits become the bottleneck. Augment Code indexes your codebase at runtime and surfaces relevant files per prompt — purpose-built for large repos but standalone. Amp inherits Sourcegraph's pre-built index, which is faster but locks you into the Sourcegraph ecosystem. If you're already on Sourcegraph, Amp is the cheaper and faster path. If you're not, Augment Code is the more flexible standalone choice. Some teams run both (Sourcegraph for code search, Augment for agentic work), but the overlap is significant enough that most pick one.

What it's like to use day-to-day. The Amp UX is competent but less refined than Cursor's. Codebase chat works well — ask 'where do we handle authentication?' and Amp returns accurate file references with cross-repo awareness when you have multiple repos indexed. Agentic mode (multi-file edits, refactors, migrations) works for medium-complexity tasks. For the deepest agentic loops — building a feature end-to-end with test-driven iteration — Claude Code is still more reliable in our testing. Amp is also newer to agentic coding than Cursor (launched mid-2025) and the feature velocity is high; the gap is closing.

Where Amp is the wrong pick: if you're a solo developer or small team where the codebase fits in a frontier model's context window, the Sourcegraph context-graph advantage doesn't compound for you — Cursor or Claude Code will be cheaper and faster. If you're on JetBrains or want a CLI-first workflow, Amp's editor support is currently VS Code-centric. If your team doesn't already pay for Sourcegraph, you're effectively paying for the index plus the agent — at which point Augment Code's standalone story is competitive.