Optic was acquired by Atlassian in April 2024 and the standalone product was archived in January 2026. We get asked the comparison anyway, often by teams who liked Optic's framing of the problem. The honest answer is that we share a primitive — capture real traffic — and solve different problems with it.
Optic shipped real engineering. The team coined "system of mistrust" to describe API contracts — the idea that a spec rots unless it is constantly tested against real traffic — and built a CLI that captured HTTP traffic in CI, generated OpenAPI from it, and diffed that spec against the spec on the previous commit. The framing was beautiful. The product was tasteful. Atlassian acquired the team in April 2024 to fold the OpenAPI expertise into Compass; the standalone product wound down over the next 18 months and the repo was archived in January 2026.
Where Gostly differs is the problem framing, not the technical quality of what Optic built. Optic was a spec-first tool: capture traffic, derive a spec, validate the spec against future commits. Useful primarily for the team producing the API. Gostly is mock-first: capture traffic, replay it as a mock, layer AI gap-fill and chaos injection on top. Useful primarily for teams consuming an upstream — internal or third-party — that they cannot easily run in dev or CI.
On monetization the two products picked different shapes. Optic ran a freemium-cloud model where the OSS CLI shipped the core spec-diff capability and Optic Cloud added collaboration and hosted lint rules. Gostly is open-core: an FSL-1.1 proxy that anyone can self-host, plus a closed cloud sidecar that handles AI gap-fill, the team workspace, and the traffic-to-training pipeline. Different bet on where the value sits and who pays for it.
| Feature | Optic | Gostly |
|---|---|---|
| Active product | Archived Jan 2026 | Yes |
| License (CLI) | MIT (archived) | FSL-1.1 |
| Captures real HTTP traffic | Yes | Yes |
| Generates an OpenAPI spec from traffic | Yes | no — different problem |
| Serves mocks from recorded traffic | No | Yes |
| AI gap-fill for unrecorded routes | No | cloud |
| Spec drift detection (CI-time) | Yes | narrow v1 (schema-diff) today |
| Live runtime drift detection | No | v1 today; canary-replay v2 Q4 2026 |
| Chaos / failure injection on recordings | No | Yes |
| Self-hostable proxy | CLI only (archived) | single binary or K8s |
Drift detection v1 (narrow) ships in cloud per PR #76. Broader live runtime drift remains roadmap. We do not claim shipping features that are not shipping today.
If you came here looking for Optic
Choose Gostly when
| Tier | Optic | Gostly |
|---|---|---|
| Free | OSS CLI, MIT (archived, final tag only) | Unlimited services, OSS proxy (FSL-1.1) |
| Pro | Optic Cloud — offline since 2024-25 | $10 / mo single user (unlimited services, AI on miss) |
| Team | n/a | / seat / mo |
| Self-host / Enterprise | n/a | $499 / mo Self-host · $25K+ Enterprise |
Optic Cloud's pricing is no longer relevant — the servers are off. We list this row for honesty: the comparison most people want to make is no longer a live one. The interesting question is whether mock-first beats spec-first for your team's actual day-to-day, not which subscription is cheaper.
If your problem is consuming an upstream you cannot easily run in dev, that is the problem we are built around. Drop in the proxy, record an hour of staging traffic, and see if the shape fits.
Evaluating for a team of 3+? We’d love to talk before you commit.