Gremlin

A commercial chaos engineering platform: agents on your hosts, containers and Kubernetes clusters take orders from Gremlin’s hosted control plane to inject resource, network and state faults, and Gremlin scores each service on how it held up.

What it is: Chaos engineering
Type: Commercial product
Made by: Gremlin Inc.
License: Proprietary
Facts checked: 24 October 2026

What it is

You install the Gremlin agent on the machines, containers and Kubernetes clusters you want to test; it authenticates with Gremlin’s hosted control plane, which you drive from a web app, a REST API or a CLI. Experiments come in three groups. Resource experiments consume CPU, memory, GPU, I/O, disk or process slots. Network experiments add latency or packet loss, drop matching traffic (blackhole), break DNS or simulate an expiring certificate. State experiments shut hosts down, shift the clock (time travel) or kill processes. If the agent misses three of its last five check-ins with the control plane, it halts its experiments and reverts them.

On top of single experiments sit Scenarios, prebuilt tests for common failures, and reliability management: you define services, Gremlin discovers their dependencies, runs a standard suite of tests for scaling, redundancy and slow or failed dependencies, and gives each service a reliability score. Health checks watch the system and halt a test if it becomes unhealthy. For serverless functions and containers where an agent can’t run, Failure Flags inject faults at the application level, either through a sidecar proxy or through SDKs for JavaScript, Python, Go, Java and .NET, on AWS Lambda, ECS, Kubernetes and others. Since October 2026 an add-on, Foresight AI, uses AI agents to suggest which tests to run, run them and propose fixes, each step needing human approval.

Why teams pick it

One product for hosts, containers, Kubernetes and serverless, with a web UI that people outside the platform team can use
Safety controls throughout: health checks that halt tests, agents that roll back when they lose contact, and company-wide halt permissions
Reliability scores and standard test suites per service, useful for tracking resilience across many teams
Failure Flags reach places host agents can’t, such as Lambda functions, and can target specific users or customers

Watch out for

Proprietary and hosted: agents must stay in contact with Gremlin’s control plane, or they halt their experiments
No published prices: the platform is priced per service and quoted after a sales call, with a 30-day trial
Shutdown and process killer experiments can’t be rolled back, so they rely on your systems recovering by themselves
Failure Flags SDKs need code changes, and the sidecar route needs traffic proxied through an extra container

Alternatives

What all chaos engineering have in common →Compare side by side →

AWS Fault Injection ServiceAWSAWS’s managed chaos engineering service: you write an experiment template of actions, targets and stop conditions, and FIS stops instances, fails over databases, slows networks or throttles AWS APIs in your account, halting when a CloudWatch alarm fires.

How it fits with other technologies

What Gremlin uses

See it in a real stack

No architecture uses Gremlin yet. Drop it into the playground and wire it up yourself.Open playground →

Worth exploring

Article· February 2024Vault chaos engineeringKalam Abdul, HashiCorpFour chaos experiments on a three-node Vault cluster with integrated storage, run with Gremlin under load from the Vault benchmark tool: losing the leader (a standby takes over with little effect), losing quorum (reads and writes stop until you follow the recovery runbook), added latency and memory pressure. A concrete worked example of hypothesis, steady-state metrics and abort conditions; written by HashiCorp with help from Gremlin, so the fixes it suggests lean on Vault Enterprise features such as performance standbys and DR clusters.

Video· April 2021Automating for Failure: Disaster Recovery Testing with Terraform & GremlinSamson Olufuwa and Olubusayo Amowe (Fuerza), published by HashiCorpA 27-minute introductory talk on disaster recovery: the common strategies (backup and restore, pilot light, warm standby, multi-region), RTO and RPO, and how keeping infrastructure in Terraform modules and AMIs lets you rebuild it quickly, with a demo on AWS and regular drills to test the plan. The slides cover Terraform far more than Gremlin, and say plainly that Terraform does not back up your data; published on HashiCorp’s site, so it favours Terraform.

Article· April 2018Automating a Chaos Engineering Environment on AWS with Terraform and GremlinHo Ming Li, GremlinA step-by-step tutorial: Terraform builds a throwaway VPC and one EC2 instance whose user data installs the Gremlin agent, you run a latency attack from the Gremlin web app and watch ping times rise by about 100 ms, then terraform destroy removes it all. Useful as a pattern for disposable chaos test environments; it is Gremlin’s own tutorial, last updated in 2018 and flagged by Gremlin as possibly out of date, and it hardcodes agent credentials in user data, which it says not to do long term.

All resources →

Where it came from2016–2026

Kolton Andrus built fault injection at Amazon in 2009 and worked as a Call Leader at Amazon and Netflix, running the response to large outages. At Netflix he was on the team that built Failure Injection Testing (FIT), introduced in October 2014 to control more precisely which components failed. He founded Gremlin in 2016, and has said he built it out of necessity after years of on-call burnout.

Gremlin came out of stealth in December 2017 with a $7.5 million Series A led by Index Ventures and Amplify Partners and early customers including Expedia, Twilio and Confluent. Its first product let teams launch attacks against hosts from agents, scheduled or at random within a time window, with an undo button. An $18 million Series B followed in 2018. The company then moved from single attacks toward measuring reliability: a reliability management platform with service scores in 2022, Failure Flags for serverless in 2023, and Foresight AI, generally available on October 7, 2026, which draws on what Gremlin calls a Failure Atlas built from more than ten years of fault-injection experiments.

Why it’s built this way

A hosted control plane with lightweight agents: Customers install only an agent; scheduling, permissions, history and reporting live in Gremlin’s service, so teams can start without building a chaos platform.
Every experiment can be halted and reverted: From launch Gremlin offered an “undo button”, and its agents revert experiments if they lose contact with the control plane, to make running faults against real systems acceptable to cautious teams.
Scores, not just attacks: Single experiments don’t tell a company whether it is getting more reliable, so Gremlin added standard test suites, dependency discovery and per-service reliability scores.

Timeline

2016
Kolton Andrus founds Gremlin
2017
Leaves stealth in December with a $7.5 million Series A
2018
$18 million Series B
2022
Reliability management platform with service scores
2023
Failure Flags for serverless and container workloads
2026
Foresight AI generally available on October 7

Sources6 · facts checked 24 October 2026

Improve this page on GitHub