Continue the conversation — chat opens pre-seeded with the current signal, caps, and movement.
Frontier agentic coding tool, now an independent company (Amp Frontier Corporation / 'Amp, Inc.', spun out of Sourcegraph Dec 2025). Differentiators: CLI-first agent surface with remote web control, automatic context compaction, subagents + Oracle, and a multi-model lineup (frontier Claude in smart mode, GPT-5.5 in deep/rush). Claims profitability (rare in the space); pay-as-you-go pricing with zero markup on provider API costs for individuals/teams.
The Feb 2026 pivot ("The Coding Agent Is Dead") removed the native IDE-extension surface in favor of CLI + web + plugins — strongest fit for teams and developers comfortable with a terminal-first, agent-centric workflow, and for those who value not being locked to one model. Teams reliant on an in-editor sidebar agent are the natural churn segment. The new company structure creates some uncertainty—enterprise commitments require confidence in vendor stability. Watch for named company-wide enterprise customers, an independent Amp-entity funding round, and Amp Labs enterprise engagements (May 2026).
Adoption & Proof Points
- Dec 2025: Spun out as the Amp entity (Amp Frontier Corporation / 'Amp, Inc.'), co-founded by Quinn Slack (CEO) and Beyang Liu (former Sourcegraph founders); Dan Adler became Sourcegraph CEO
- Claimed profitability (rare in AI coding space)
- May 27 2026: Neo CLI rebuild shipped to everyone (remote control, auto-compaction, plugin API)
- Pricing: pay-as-you-go, zero markup on provider API costs ($5 min credit) for individuals/teams; Enterprise = $1,000 one-time activation + 50% markup on usage
- May 2026: Amp Labs — embedded CEO-mandate enterprise engagements (profit via stock warrants; one company per industry/region)
- Developer testimonials from Cloudflare (Glen Maddern, Principal Systems Engineer), Instacart engineers
- Backed by Craft, Redpoint, Sequoia, Goldcrest, a16z (carried from Sourcegraph; ~$178M); no independent Amp-entity round announced post-spinout
- Gap: No Fortune 500 / named company-wide deployments documented
Recommended Use Cases
- Teams and developers comfortable with a terminal-first, agent-centric workflow (CLI + web + plugins)
- Organizations valuing thread sharing, remote agent control, and persistent context across developers
- Teams that want model flexibility (frontier Claude in smart mode, GPT-5.5 in deep/rush) rather than single-model lock-in
- Cost-sensitive teams that prefer zero-markup pay-as-you-go over per-seat subscriptions
- NOT a fit for teams that require a native in-editor sidebar agent (extensions discontinued Feb/Mar 2026)
Risks & Limitations
- CLI-only pivot: Native VS Code/Cursor editor extensions discontinued (Feb/Mar 2026). Teams reliant on an in-editor agent sidebar must adopt the CLI (+ thin `amp --ide` bridge) or look elsewhere; competitors (Kilo etc.) are actively courting displaced IDE-extension users
- New company structure: Dec 2025 spinoff (Amp Frontier Corporation) creates vendor-stability questions for multi-year enterprise commitments; no independent Amp-entity funding round announced post-spinout
- ISO 27001 gap: SOC 2 Type II is held by the Amp entity, but ISO 27001 is NOT (remains under Sourcegraph only)
- No self-hosted option: Cloud-only (GCP US) may exclude regulated industries requiring on-premises/air-gap
- Pricing predictability: Public but usage-based with zero markup (individuals/teams) and a $1K-upfront + 50%-markup Enterprise model — no per-seat list price or volume tiers, challenging enterprise budgeting
- No Fortune 500 validation: Individual developer testimonials (Cloudflare, Instacart) but no company-wide enterprise deployments documented
Capabilities & Integration
Agentic depth: Subagent orchestration for parallel task execution and context preservation. Oracle 'second opinion' tool (GPT-5.5, high reasoning) for complex reasoning/analysis. Smart mode runs frontier Claude (Opus 4.7 as of late May 2026; expected to track Opus 4.8 — released 2026-05-28); deep mode runs GPT-5.5 with extended thinking; rush mode runs low-token GPT-5.5 with no reasoning for small tasks. Multi-agent workflows for large-scale refactoring.
Context handling: Automatic context compaction (Neo) triggers at ~90% of the context window so threads no longer die at token limits. 1M-token window (Sonnet default). Persistent threads stored server-side, resumable anytime; remote control of running CLI agents from ampcode.com (live updates, message queuing/dequeuing, cancellation). Librarian subagent searches public and private GitHub repos. AGENTS.md files (CLAUDE.md compatible) for cross-tool rules. (Thread handoff was removed and replaced by auto-compaction in the Neo rebuild.)
Interface surface (post Feb 2026 pivot): CLI-first (rebuilt 'Neo', npm package now @ampcode/cli) + web (ampcode.com threads + remote agent control) + Plugin API (web UI + terminal sync, 'Plugins, Everywhere' May 28 2026) + thin `amp --ide` bridges that give the CLI access to the active file/selection/diagnostics in VS Code and Neovim. Native VS Code/Cursor editor extensions were discontinued ("The Coding Agent Is Dead," Feb 19 2026; self-destruct Mar 5). MCP support with Enterprise registry allowlists.
Extensibility: Plugin API for tools/commands/UI/reasoning without full MCP servers. Public/private/workspace thread visibility with user-group controls.