What is Keploy?
Keploy turns traffic into tests. You start your service under keploy record, use it, and Keploy captures the incoming API calls together with the database queries, queue messages and outbound calls they caused; keploy test then replays the requests and serves the recorded dependencies as mocks, so the suite runs offline with no test database to set up. Mocked dependencies include PostgreSQL, MySQL, MongoDB, Redis, Kafka and RabbitMQ, over HTTP, gRPC or GraphQL.
On Linux it captures at the network layer with eBPF, which is why it needs no SDK or code change but does need root and kernel 5.10 or later. Native macOS (Apple Silicon only) and Windows builds intercept in user space and only replay HTTP, MySQL and MongoDB reliably; other services need Docker. AI features generate API tests from OpenAPI or Postman specs and expand edge cases such as boundary values, missing fields and wrong types.
The Go repository has about 19,500 stars and is Apache-2.0, with releases several times a week (v3.6.222 on 6 November 2026). Recording and replay require a free Keploy account.
Is Keploy worth it?
Integration tests are expensive mainly because of their dependencies: a seeded database, a Kafka broker, a stub for the payment API. Keploy's answer is to record what those dependencies actually returned and replay it. For a service with no tests, an afternoon of clicking through flows under keploy record can produce a suite that runs in CI without any infrastructure, which is a better starting point than the usual blank folder.
The eBPF approach is what makes it language-agnostic, and also what makes it intrusive. On Linux, record and test run as root inside the kernel; rootless Podman cannot grant eBPF at all, and low-latency mode on arm64 needs kernel 6.4. Security teams should treat it like any other privileged agent and keep it out of production hosts unless they mean it, which is also where Enterprise sells production traffic capture.
The catch: recorded tests only cover what someone did while recording, and recorded responses go stale when the schema changes. The AI edge-case generation on paid plans helps with the first problem; the second still needs re-recording. On macOS and Windows the native builds replay only HTTP, MySQL and MongoDB, so a Postgres or Redis service has to run in Docker. A free account is now required even for the open-source CLI, and the repository carries 385 open issues.
The pricing is modest. Pro at $19 a user a month includes $19 of usage, and 400 runs a month cost nothing extra; beyond that each test run is $0.22. Security pages show SOC 2 Type II and ISO 27001, though SSO and audit logs are Enterprise-only.
For backend teams with thin test coverage and a Linux CI runner, Keploy delivers more per hour invested than writing mocks by hand.
Where it earns its place
A four-person backend team running a Go and PostgreSQL billing service
Get regression coverage before a risky refactor
Run the service under keploy record on a Linux dev box, exercise the main flows, commit the YAML tests and mocks, and run keploy test in GitHub Actions; the Postgres calls are served from recordings, so no database container is needed.
A QA lead at a fintech with a published OpenAPI spec
Cover edge cases nobody writes by hand
Generate API tests from the spec on Playground or Pro and let the AI add boundary values, missing fields and wrong types; Pro's 100 generated suites a month suit a handful of services, with $0.16 per extra generation.
A platform engineer evaluating the tool for 30 microservices
Decide whether it can run in Kubernetes staging
Pro adds Kubernetes capture in dev, QA and staging clusters; the DaemonSet agent's low-latency mode needs x86-64 kernel 5.10 or arm64 6.4, otherwise use sidecar mode.
Before you commit
- Setup
- Install with the curl script, then keploy record -c "<command to start your app>" and keploy test; a free Keploy account is needed for both.
- Integrations
- Mocks PostgreSQL, MySQL, MongoDB, Redis, Kafka and RabbitMQ over HTTP, gRPC and GraphQL; runs in Jenkins, GitHub Actions and Kubernetes; imports OpenAPI and Postman specs; OpenAI and Gemini models with bring-your-own-key.
- Data and privacy
- Recordings, including database responses, are stored as local YAML in the open-source CLI; Playground and Pro are cloud-only; Enterprise can be self-hosted or air-gapped. The pricing page shows SOC 2 Type II, ISO 27001, GDPR and HIPAA badges.
- Cost at scale
- Pro is $19 per user a month with $19 of usage, 400 test runs and 20,000 sandbox runs included, then $0.22 per run and $10 per 10,000 sandbox runs; SSO, audit logs and production capture need Enterprise at a custom price.
- Limits
- Linux capture needs root and kernel 5.10+; native macOS is Apple Silicon only; tests cover only what was recorded.
What’s good
- Captures API calls and their database, queue and HTTP dependencies with no code change
- Replays offline with recorded mocks, so CI needs no test infrastructure
- Works across languages because capture is at the network layer
- Apache-2.0 with very frequent releases
- Pro from $19 a user with usage included
What’s not
- Needs root and eBPF on Linux; rootless Podman unsupported
- Native macOS and Windows replay only HTTP, MySQL and MongoDB
- Recorded mocks go stale when schemas change
- Free account required even for the open-source CLI
Best for
Backend teams with little test coverage who want API and integration tests without building mocks or test databases.
Keploy alternatives
Keploy: common questions
- Is Keploy worth it?
- Keploy is the quickest route from an untested backend to a replayable API suite with real dependency mocks, as long as you can grant it root on Linux. We rate it 4.0 out of 5.
- How much does Keploy cost?
- Open-source CLI free; Playground free (30 generated suites, 100 test runs a month); Pro $19/user/month including $19 of usage, then $0.16 per test generation, $0.22 per test run, $10 per 10,000 sandbox runs; Enterprise custom.
- What are the downsides of Keploy?
- Needs root and eBPF on Linux; rootless Podman unsupported; native macOS and Windows replay only HTTP, MySQL and MongoDB; recorded mocks go stale when schemas change; free account required even for the open-source CLI.
- Who is Keploy best for?
- Backend teams with little test coverage who want API and integration tests without building mocks or test databases.
- How much does Keploy cost for a team or in production?
- Pro is $19 per user a month with $19 of usage, 400 test runs and 20,000 sandbox runs included, then $0.22 per run and $10 per 10,000 sandbox runs; SSO, audit logs and production capture need Enterprise at a custom price.
- Is Keploy safe for company data?
- Recordings, including database responses, are stored as local YAML in the open-source CLI; Playground and Pro are cloud-only; Enterprise can be self-hosted or air-gapped. The pricing page shows SOC 2 Type II, ISO 27001, GDPR and HIPAA badges.
- What are the best Keploy alternatives?
- The closest alternatives we have reviewed are Playwright Test Agents and promptfoo.
Answers checked Nov 6, 2026.
Make Keploy? Show this rating on your site
