EAS by Lynxmind

Run your orders, connect your systems, leave legacy integration behind

EAS is one platform for three jobs. It discovers what already runs, turns a sentence into a tested automation, and shows every request end to end, on your Kubernetes, in the cloud or on-premises.

  • Runs on your Kubernetes
  • Native, governed AI
  • Migrates from TIBCO, MuleSoft, Boomi and more
fibre-activation

Live trace

You: Activate a fibre order: take the payment, provision the line, wait for the router, then start billing.

  1. Stagepayment.capture
    0.3 s
  2. Stagenetwork.provision
    1.8 s
  3. Wait, SLA 2 hrouter.activated
    4 min
  4. Stagebilling.start
    0.1 s

Order activated

FIB-1042

Three ways in

One platform, three jobs

Start where it hurts most. Order Management, integration and migration share one catalog, one login and one identifier through all of it.

Order Management

Every order, visible from intake to completion

Business processes with automatic stages, waits with an SLA and team tasks, without BPMN or code. The back office works in a console of its own; the processes live in EAS.

Order journey
One order number, its whole path: where it is, where it failed, how long it took.
Waits and tasks
Waits with an SLA, team tasks with a form, notices by email and Teams.
Safe cancellation
Cancelling half-way undoes what was done, in reverse order.
ITSM included
Tickets, changes, problems, CIs, SLAs and defects, linked to the automations.

Ask the Assistant

  • “Why is order 1042 stuck?”

    A ticket opens with the cause analysed by the AI, in under 15 minutes.

  • “Give me a dashboard of today's orders”

    Running, waiting, late and completed, proposed from the data it knows.

See it on your case

Integrator

From the catalog to the canvas, without one more adapter

EAS fills an executable catalog from your cluster. Automations are designed on a canvas or in a sentence, tested against real data, and promoted as the same image to Production.

Saga Studio
Blocks, breakpoints, a live trace with real traffic, versions side by side.
Data Mapper
Auto map and 80 functions for dates, lists, text, maths and regex.
Assistant
Design, copy, run, import, repair and explain, by sentences.
Triggers of your choice
API, schedule, Kafka, an event from another saga, or the chat itself.

Ask the Assistant

  • “Convert this XML to JSON”

    XML, JSON and XSD conversion, right in the chat.

  • “Generate a microservice that…”

    A Ballerina service, built and tested in Docker, lands in the catalog.

See it on your case

Migrator

From your current integration platform, export by export

An export of TIBCO, Apigee, MuleSoft, Boomi, webMethods, SAP Integration Suite, Camel, Logic Apps, IBM App Connect and more becomes catalog entries and a saga.

One profile per technology
A new technology is a prompt you edit, not a new release.
Checked, not guessed
What the AI read is checked against the files; inferred items stay marked.
Proven before the switch
Regression runs and QA sign-off per version before promotion.
Breakpoints as in BusinessWorks
Debug the installed version, without touching production.

Ask the Assistant

  • “Catalog the operations in this TIBCO repository”

    A source profile reads the code and catalogs the operations with their schemas.

  • “Import this Git folder of sagas”

    One folder, or a batch of them.

See it on your case

How it works

Everything that runs, described by itself

EAS asks Kubernetes which services exist, and each one hands over its contract. Files, repositories, documentation and platform exports feed the same catalog, and everything else in the platform reads from it.

  • Not only for design

    The catalog is the living inventory behind telemetry, tenant management and the Assistant's answers.

  • Version, folder and owner

    Folders grant access by group, and the Impact Map says who depends on what.

  • One action, every environment

    Endpoints written with ${VARIABLE} call the Development CRM here and the Production CRM there.

Governed AI

One Assistant for everyone, with the AI governed in your hands

Every request goes through the same gates. The answer is a proposal to confirm, and who confirmed it is recorded.

  • The provider is yours

    A self-hosted model, OpenAI, Anthropic, Gemini or any compatible API, per environment and per purpose.

  • Prompts you can edit

    System prompts and personas in AI Config, with versions; no prompt hidden in code.

  • Every call is logged

    Purpose, tokens, duration and model; the Costs page shows spend by day, model and saga.

system prompts
36
personas
18
registered tools
91

Why is order 1042 stuck?

Context builder

  1. Router

    Picks the domain and the persona

  2. Guardrails

    Never secrets or clear data

  3. Persona

    Tuned per tenant

  4. Knowledge

    Your docs and the platform

  5. Tools

    Only what the registry allows

Proposal

erp-adapter has had no healthy replica since 14:03. Scale it from 1 to 2 and retry the block.

Nothing happens without a yes.

Role by role

What changes for each team

Underneath the three jobs is one platform. Pick a role to see what it gets.

Developers

From the catalog to the canvas, without one more adapter

  • Drag real operations from the catalog; the schemas come with them.

  • Sagas compiled into native images, each running in its own pod.

  • Comments per version, so the review stays next to the change.

  • Pro-Code when you want code: VS Code in the browser, with the Assistant inside.

Testers

An API changed. See what breaks before production

  • Impact Map: the automations that depend on an action, down to the block.

  • Regression runs, for real or simulated, with each one's last real payload.

  • Defects recorded on the version and the run where they were seen.

  • Saga health: broken latest versions flagged every hour.

Delivery

The same image, from Development to Production

  • Publish produces an immutable image; promotion takes the same digest onward.

  • Environments as a pipeline, each with its own cluster and tier.

  • Release approvals where there is a lock, with a record of who approved.

  • Drift is visible, and roll back is one click.

Operations

One identifier end to end, and the cause before the phone rings

  • Every request has its timeline, service by service.

  • Dead letters and reprocessing, with retry policies by error code.

  • Kafka watched: brokers, lag per consumer, under-replicated partitions.

  • Grafana with Alloy, Mimir, Loki and Tempo, installed from the start.

Security

Each tenant in its place, every refusal with a name

  • Tenant isolation in every table and query, with a data key per tenant.

  • Company Microsoft login, by invitation only.

  • Emails, tax numbers, IBANs, cards and phones leave the logs masked.

  • Credentials encrypted with AES-256-GCM; ISO 27001 documentation under way.

Architects

Independent services, one entry point, everything in git

  • 23 independent Ballerina services, each scaling on its own.

  • One gateway with JWT, role and tenant on every request.

  • Everything in git, with encrypted secrets and backups with a tested restore.

  • Built-in connectors: Microsoft 365, GitHub, Kafka, JMS, SMTP and PostgreSQL.

FAQ

Questions before a walkthrough

Where does EAS run?

On your Kubernetes: k3s on-premises, AKS or EKS. One installer sets it up in 1 to 1.5 hours, and the whole platform can be rebuilt from git on an empty cluster in 6 to 7 minutes.

Do we have to replace our current integration platform at once?

No. The Migrator brings integrations in export by export. TIBCO, MuleSoft, Boomi, webMethods, SAP Integration Suite and more become catalog entries and sagas, proven by regression runs before anything is promoted.

Which core components are open source?

The open-source core components include Kafka (KRaft), PostgreSQL with pgvector and Karapace. EAS also has 19 modules that can be enabled per client.

How is AI governed?

Your team chooses the provider (self-hosted, OpenAI, Anthropic, Gemini or any compatible API) per environment and purpose. Prompts and personas are versioned, every call is logged, and the AI never receives secrets or clear data.

Get started

See it run on your case

We show the platform running on your case: your cluster discovered, an automation designed in front of everyone, and the path of one request end to end.

Tell us which job comes first: your orders, a new integration, or a migration away from your current platform.

Book a walkthrough

Prefer to write? Send us a message at info@lynxmind.com

TIBCO, MuleSoft, Boomi, webMethods, SAP, Apigee, IBM App Connect, Logic Apps, Kubernetes and other names are trademarks of their respective owners.