Latest newsTalend 7.3: extended support ends in December 2026. Our exit plan
The data integration runtime you own
Your flows, written by Claude, GPT, Gemini, your own AI agent or yourself.
Receive, route, transform and deliver your data between your applications, on your own servers, with a platform that belongs to you.
What is Portlane?
Your teams specify, AI writes the flows in Python, and it all runs on your servers, supervised from one screen.
- On your serversYour data centre or your private cloud. No data goes through us.
- In PythonFlows your team reads, tests and versions in Git.
- ResilienceEvery delivery is kept, traced and replayable in one click.
- No meterA licence paid once. Nothing per message, per volume or per bandwidth.
Building a flow has changed
A new way of working, not a new tool.
Yesterday: a ticket, a studio, a queue.
- The businessStates the need, in plain words.
- The AI agentWrites the flow and its tests.
- The data engineerReviews, approves, ships.
Ready to take the agentic turn?
Make your data flows, the integrations between your business applications and that part of your information system agentic.
- Python, the main language of AI and the backbone of Portlane, though not the only one.
- Portlane AI Ready: an MCP server hosted on your instance gives your agent the platform's documentation and tools.
- Twenty years of integration patterns, built into the core.
Claude, GPT, your LLM… Portlane welcomes coding agents, with no technical barrier to entry.
The need, written by the business
- Every new CRM contact becomes an ERP customer.
- The email is stored in lower case.
- With no country given, France is the default.
The tests, written by the agent
def test_le_contact_devient_un_client(): assert sortie["code_client"] == "C-10482"def test_l_email_est_en_minuscules(): assert sortie["email"] == "j.martin@exemple.fr"def test_sans_pays_la_france(): assert sortie["pays"] == "FR"Unlimited flows, one mechanism
Portlane gives your data a robust way to travel: everything links up, everything replays, nothing is lost.
- Nothing is lostEvery record is put away in your database before any processing. If an application goes down, it waits.Sends the contactKeeps it storedDown, then back
CRM : Sends the contact. Portlane : Keeps it stored. ERP : Down, then back. - Never twiceSending a record again does not create the same invoice, or the same customer, twice.Sends it twiceDrops the duplicateGets it once
CRM : Sends it twice. Portlane : Drops the duplicate. ERP : Gets it once. - Everything replaysA delivery fails? Once the cause is fixed, one click sends it again.Sends the contactReplays it in a clickRefuses, then accepts
CRM : Sends the contact. Portlane : Replays it in a click. ERP : Refuses, then accepts.
The Control Tower, on your servers
Every flow, every delivery, every failure, on one screen and on your instance. Try it right here.
Watch the demo- Instance
- inst-7c1e42
- Environnement
- Production
- Cœur
- v1.2.1
- Licence
- Valide
- Dernière lecture
- 14:32:07
- Instance
- inst-7c1e42
- Environnement
- Production
- Cœur
- v1.2.1
- Licence
- Valide
- Dernière lecture
- 14:32:07
Vue 360°
Ce qui tourne, ce qui attend, ce qui demande un regard.
À examiner
2- Cible supprimée14 enveloppes en attente
- 2 lettres mortesEnveloppe 4 · Enveloppe 5 · à rejouer
Activité
3 dernières- 14:31:57Livrée à erpEnveloppe 1
- 14:31:57TransforméeEnveloppe 1
- 14:31:57RoutéeEnveloppe 1
309 msfrom the CRM to the ERP
Portlane times every step: the Tour shows you where every millisecond goes.
Portlane takes 125 ms of them. The rest is the ERP answering.
- 0 msCRM
Sends the new contact.
- 125 msPortlane
Receives it, stores it, routes it and transforms it.
- 184 msERP
Receives it and answers 200.
Leave your current tool, or complement it
Portlane takes over your HTTP/JSON flows one at a time, the old tool running alongside until the switch.
Exit diagnostic, 30 minA new view of the integration tools market (ETL, iPaaS…)
The job of an iPaaS or an ESB, not of an ETL, is applications in real time.
AI is shaking the hold of low-code tools: Portlane is built to make working with AI agents easy.
See all comparisons- Map of the integration market: where each tool sits
- Data pipelines in code : Dagster, dbt, Prefect, Airbyte, Snowflake
- Application integration in code : Portlane
- Visual ETL and ELT : Talend, Fivetran, Informatica, Fluhoms
- Studio iPaaS and ESB : MuleSoft, Tibco, Boomi, Blueway, Workato, n8n, Zapier, Make
Price
- Installed ETL
- Licences and maintenance
- Portlane
- Lifetime license, on quote
Flow format
- Installed ETL
- Proprietary studio
- Portlane
- Python code in Git
With an AI agent
- Installed ETL
- Visual studio, hard for an agent to read
- Portlane
- Claude, GPT, your LLM… Portlane provides the skills
Where it runs
- Installed ETL
- On your side
- Portlane
- Your servers
Exit cost
- Installed ETL
- Scarce skill, closed format
- Portlane
- You keep the software
Connectors
- Installed ETL
- Large historical catalog
- Portlane
- Few, chosen: HTTP/JSON
Governance
- Installed ETL
- Depends on the suite
- Portlane
- A pure data-exchange player: no dedicated MDM feature
Your data stays with you
No business data goes through a Portlane cloud: everything runs and is stored on your servers.
- 0 bytes to Portlane
- PostgreSQL on your side
No open door
Your servers fetch the updates. Portlane has no inbound access and no SSH, and installing can be done without internet.
- Pull only
- Offline bundle
Every version can be checked
Every image is signed and ships with the list of its components: your teams check it before installing.
- cosign signature
- SBOM
- sha256 checksums
How much does Portlane cost?
A lifetime licence, on quote. Its price depends on your scope and your deployment, never on volume or bandwidth.

Oncelifetime licence, on quote
Zero euros per message
- Unlimited flows within the licensed scope.
- On your side: your servers, a VPS or your private cloud.
- New versions included. Support by the hour, when you need it.
Frequently asked questions
Is Portlane an iPaaS?
It does the job, not the business model. Like an iPaaS or an ESB, Portlane connects your applications in real time. But it runs on your side, with a fixed license: your flows and Postgres stay in your perimeter.
Where does Portlane run?
Only on your side: your servers, data center, or private cloud. Portlane and Postgres stay in your perimeter. Nothing routes through a Portlane cloud.
How are flows written?
Your team specifies them; an AI agent like Claude or GPT writes them in Python, with their tests. They are versioned in your Git repository: review, CI, and rollback like the rest of your IT code. No proprietary visual designer.
How much does a Portlane license cost?
On quote. Lifetime license, fixed cost based on scope and deployment. No usage pricing and no public self-serve price list.
Does Portlane replace Talend or Blueway?
For the core receive / transform / deliver over HTTP/JSON, yes for many mid-market teams. No BPM, MDM, complex EDI, or certified connector catalog — the scope is deliberately bounded.
Does Portlane read SQL databases or files?
No. The shipped scope is HTTP/JSON: signed webhooks and scheduled API calls. Few connectors, by choice: performance, maintenance and security come before the catalog.
Are new versions included?
Yes. The license gives access to new versions, with no yearly fee; you install them at your own pace, under your own responsibility. Support is separate, as a bundle of hours used on demand. An assisted upgrade is a project quoted on its own.
Future-proof
Our philosophy: tools that last.
- Born with AI, with open arms.Portlane welcomes agents to make the platform stronger, and the flows of your information system resilient.
- AI within a frame, not in a heap.The core is built on twenty years of integration patterns: AI stays maintainable, instead of a tangle of flows and tools.

Stop renting your integration layer.
Thirty minutes with our engineering team: your scope, your installation, your quote.




How it works
Specify, your agent writes
The business describes the need; the AI agent writes the flow and its tests.