The problem it solves
A single business process now runs across several teams, legacy systems, new services and, more and more, AI agents. Work gets lost in the handoffs between them:
- cases stall with no owner somewhere between one team and the next;
- no one can say, right now, where each case is and why it stopped;
- approvals, sign-off limits, deadlines and reminders are tracked in email, spreadsheets or people's memory;
- integrations fail halfway through and force you to start over;
- incidents are investigated across scattered logs;
- AI is used in operations with no limits, no human review and no record of what it did.
Kikwiflow addresses the common root cause: there's no layer coordinating the work end to end.
How it works
Three components, one for each stage of the process.
Craft: model
You design how the work should happen: who does what, which systems are involved, where decisions are made, when to wait and what to do if something fails. The model isn't side documentation. It's the actual process that will run, saved as a .kikwi file (JSON), separate from your code.
Engine: run
Engine runs the process and stores the state of every case. Whether it has to wait for a person, retry, request approval, pause or resume, the process knows exactly where it is. Technically, it's a Java 21 library that runs inside Spring Boot applications, with persistence in MongoDB.
Monitor: observe and intervene
Monitor shows every case in flight: which step it's on, who or what performed each step, where it stopped and what comes next. From there your team can fix, approve, retry or resume, without reconstructing the story from logs.
Who it's for
- Decision-makers get real-time operational context: where cases are, how long they sit at each step and where people need to step in.
- Operators stop dealing with a black box: when a case gets stuck, they see the exact step and resume from there.
- Builders discuss business rules in the same artifact that will run, and keep the technical implementation in code.
When to use Kikwiflow
- The process runs through people, multiple systems and AI, and no one knows where each case stands.
- You want to put AI agents into operations with limits, human approval and an audit trail.
- Approvals, SLAs, waits and escalations are managed by hand.
- Integrations between legacy and new systems fail midway and work starts over from scratch.
- Orchestration has to run on your own infrastructure, with data under your own policies, for example to meet data-residency requirements.
When Kikwiflow isn't the best fit
- Pure data pipelines and analytics workflows.
- Kubernetes-native batch job scheduling.
- Generic automations between SaaS apps and simple integrations.
- Enterprise process management where the BPMN standard is a requirement.
- Pure durable execution, where orchestration lives entirely in code.
Those scenarios have tools designed specifically for the job. See our comparisons with Temporal and Camunda.
Open core model
The Community Edition is free, licensed under Apache 2.0 and runs on your infrastructure with no sign-up. For business-critical operations, atoxfy, the company that builds Kikwiflow, offers the Enterprise plan: dedicated 24/7 support, consulting with the people who built the engine, early access to the roadmap and training.
Is it already in production?
Yes. Since September 2025, Kikwiflow has orchestrated the CRM of a large automotive sales franchise network 24/7, coordinating more than 45,000 deals. See the CRM use case.
How to get started
- Model your first process in Craft.
- Follow the getting started guide in the docs to run it inside your application.
- If the process is business-critical, talk to an atoxfy expert.
Frequently asked questions
Is Kikwiflow free?
Yes. The Community Edition is free, licensed under Apache 2.0, runs on your own infrastructure and requires no sign-up. For mission-critical operations, atoxfy offers the Enterprise plan, with dedicated 24x7 support, consulting and training, priced on request.
Who builds Kikwiflow?
atoxfy, a Brazilian company that builds the engine and provides support, consulting and training for teams running Kikwiflow in mission-critical operations.
What technology does Kikwiflow run on?
The Engine is a Java 21 library that runs inside Spring Boot applications, with MongoDB persistence. Processes are modeled in Craft and saved as .kikwi (JSON) files, kept separate from the code.
Does Kikwiflow replace the systems a company already uses?
No. It coordinates the systems, people and AI agents that are already part of the operation, including legacy systems. Each system keeps doing what it does today and becomes a step in the process.