Skip to content

AI that does real work on your systems, by your rules

Reelease understands and documents your systems from their code. Your team publishes what each agent can run, and the agent only takes those steps, each with its own rules. By default, changes wait for another person, and everything goes into a run record.

We email you the link to open on your computer.
Read the transcript

The safe way to put AI to work on critical systems and in regulated sectors.

Here, a mistake is a transfer that never goes out or a missed deadline.

Laura asks for what she needs and the agent proposes a plan.

Reelease checks the rules and the permissions…

…and signs the plan: if anyone alters it, it doesn’t run.

The lookups go ahead on their own.

The change waits for someone else: whoever starts the task can’t approve it.

María approves and the transfer goes out, on the same core as always.

Resolved within the shift, with nothing rewritten.

Every task leaves its run record: who asked, what was done and who approved it.

And you can stop an agent instantly.

On the systems you already run. No migration, no rewrite. And it keeps working when you replace them.

And how does the agent know what to do?

First, Reelease has read your system, exactly as it is.

It reads the code and the documentation and builds a map your team reviews.

And it leaves everything documented, with its sources, as PDF or Word.

Your team chooses which use cases each agent can execute.

Only what your team publishes can run. Everything else stays documented, for reference.

Every agent starts with no permissions: an admin turns on only what it needs.

Your AI, where you decide.

It runs where you decide, with the AI model you choose.

Your data is looked up, never copied. Nothing is used to train models.

For sectors that can’t improvise: the same control, with concrete cases in every sector.

Reelease, the control layer between AI and your systems.

AI for critical systems. Under control.

Reelease is the control layer between AI and your systems: its agents work on what you already run and do only what your team publishes, you decide which steps a person approves, and every run leaves a run record, the complete audit trail of the task.

One executable use case combines the core, your tools, AI steps and people, so you can improve the process too.

Once AI starts to act, control has to live inside every run.

What it is
  • Your system understood from its code: its map, its documentation and its use cases, with their rules, screens and data, each with its source. Your team reviews them and turns the ones it chooses into tasks for your agents.
  • Agents that work on what you already run, from the mainframe to your email.
  • Deterministic: it runs the Sealed Plan you signed, task by task, and your rules check every step first. With Reelease’s own engine, the default for plans with a connector, a step with all its data is compiled from the Sealed Plan: the model doesn’t write the call. Otherwise the call stays pinned to the sealed step, and changes are approved on their exact values. It does what you sign off on. Nothing more.
  • Access per person and per agent, starting from zero.
  • No hallucinations: a chat that answers with its source, in your code or your documents, checks each citation against it and tells you when there is none. It also drafts with you.
  • Changes a person approves by default, and a record of every run.
What it isn’t
  • A wiki the agent reads without anyone reviewing it, or documentation that sits in a drawer.
  • A migration or rewrite project, like those from integrators, IBM watsonx Code Assistant for Z or AWS Mainframe Modernization.
  • An agent that does whatever its natural-language instructions say. You talk to Reelease the same way, but what you write becomes a plan your rules check before it runs.
  • An agent with access to your whole core and no rules.
  • A generic AI chat, such as ChatGPT, Claude or Gemini.
  • A coding copilot that proposes changes nobody controls.
  • A bot that automates screens (RPA), such as UiPath, Automation Anywhere or Blue Prism.
  • Agentic process automation, such as Maisa or UiPath Agentic Automation.
  • A stable layer between your agents and your systems: when you replace one, you change its connection and the agents carry on as before.
  • Whatever the analysis finds stays a proposal until you publish it.

Brands named are examples of each category and belong to their owners. Their presence does not imply affiliation or endorsement.

AI agent glossaryAudit trail, guardrails, human-in-the-loop: what each term means and what Reelease calls it.See the glossary

Compared with what you’re already evaluating

Migrating, automating screens, an AI chat or process agents. Reelease understands your system from its code and works on the core under your control, in Europe.

  • How it understands your system

    Reelease

    Yes: From your code

    Reads the code and documentation, from Oracle Forms and Reports to your APIs, and maps your system and its use cases, with the source of every fact

    The alternatives

    Migrate or rewrite

    Partly: Through an analysis at the start of the project, to rewrite it

    RPA

    No: From what it sees on screen

    Generic AI (chat)

    No: From what you tell it in the conversation

    Agentic process automation

    Partly: From the process as it is described and configured

  • Work on your core

    Reelease

    Yes: On the core

    Mainframe and CICS, PL/SQL and databases inside your network. Only the executable use cases you publish act on it

    The alternatives

    Migrate or rewrite

    No: It replaces it: everything is rewritten and moved

    RPA

    Partly: Through the screen, and it breaks when the screen changes

    Generic AI (chat)

    No: It doesn’t touch it: someone acts by hand afterwards

    Agentic process automation

    Partly: Mostly back-office processes and documents

  • What runs

    Reelease

    Yes: Deterministic

    The Sealed Plan Reelease builds and you sign, task by task, and your rules validate every step. With Reelease’s own engine, the default for plans with a connector, the model doesn’t write the calls. If the plan changes, it doesn’t run

    The alternatives

    Migrate or rewrite

    Partly: The new code, once the project ends

    RPA

    Partly: A fixed script of clicks and keystrokes

    Generic AI (chat)

    No: Nothing: it only answers

    Agentic process automation

    Partly: The configured process, with its rules and exceptions

  • Human approval

    Reelease

    Yes: You approve every change

    With Two-Key Approval, your rules set who approves (someone else, by default). You approve the exact values; if they change, you’re asked again

    The alternatives

    Migrate or rewrite

    No: The project is approved, not each change

    RPA

    Partly: Added by whoever designs the bot, if they plan for it

    Generic AI (chat)

    No: The person decides what to do with the answer

    Agentic process automation

    Partly: Human review at the points defined when the process is configured

  • Answers about your system

    Reelease

    Yes: No hallucinations

    Every fact about your system comes with its source, in your code or your documents, and each citation is checked against it; when there is none, it says so. No invented action reaches your systems

    The alternatives

    Migrate or rewrite

    Partly: Whatever documentation the project leaves

    RPA

    No: It doesn’t answer questions

    Generic AI (chat)

    Partly: From what the model knows, with no source in your code

    Agentic process automation

    Partly: From the process knowledge it was given

  • Proof of every run

    Reelease

    Yes: A run record for every run

    Every step, who approved it and which account talked to each system. Exportable, and you can replay the run from it

    The alternatives

    Migrate or rewrite

    No: The project’s documentation, not each task’s

    RPA

    Partly: The bot’s logs

    Generic AI (chat)

    No: A chat thread, with no shared record

    Agentic process automation

    Partly: A log of the process steps

  • Where your data lives

    Reelease

    Yes: In the EU

    Service in Madrid and models inside the European Union, or in your cloud or on your premises

    The alternatives

    Migrate or rewrite

    Partly: Wherever the project contract says

    RPA

    Partly: Wherever the bots are deployed

    Generic AI (chat)

    Partly: Wherever the chat service runs

    Agentic process automation

    Partly: In the region the vendor offers

  • When you see value

    Reelease

    Yes: From the first use case

    With your first executable use case published, on what you already run

    The alternatives

    Migrate or rewrite

    No: At the end of a long project

    RPA

    Partly: Soon, on screen tasks

    Generic AI (chat)

    Partly: Right away, but only as answers

    Agentic process automation

    Partly: Soon, on back-office processes

  • When you replace a system

    Reelease

    Yes: No lock-in

    You switch the connection; agents, approvals and run records stay the same

    The alternatives

    Migrate or rewrite

    No: It is the replacement: all at once

    RPA

    No: The bots have to be rebuilt

    Generic AI (chat)

    No: Nothing changes: it doesn’t run anything

    Agentic process automation

    Partly: It depends on how each process was connected

  • Migrate or rewrite:
    Integrators, IBM watsonx Code Assistant for Z, AWS Mainframe Modernization
    RPA:
    UiPath, Automation Anywhere, Blue Prism
    Generic AI (chat):
    ChatGPT, Claude, Gemini
    Agentic process automation:
    Maisa, UiPath Agentic Automation

Brands named are examples of each category and belong to their owners. The table compares categories, not specific products.

Frequently asked questions

What is Reelease?

The control layer between AI and your systems: a platform that understands and documents what you already have, draws out its use cases and puts AI agents to work on them. They run only what your team publishes, with the approvals you define and a record of every run.

Is it a copilot or a chatbot?

No. A copilot answers in a chat. Reelease agents act on real systems: they check their rules before starting, get human-in-the-loop approval on the steps you choose and record what they do.

Can we improve a process as well as run it?

Yes. Building on what Reelease understands about the process, you can add steps that are done by hand today (an email, a ticket, a lookup in another system), combine data from several systems in the same task and replace pieces one at a time. Each new step has its own rules and approvals and is part of the run record.

What’s the difference between AI agent governance and security?

Governance decides what agents can do and who can do what: rules, executable use cases, approvals and per-person access. Security protects the environment: each operator’s identity, least privilege by default, encrypted credentials, Sealed Plans and records.

How do we get started?

Tell us about a process or a system that matters to you. We’ll show you Reelease on a similar case and, if it fits, run a real 3-month pilot on your own systems with you.

Who is Reelease for?

Companies and public bodies with critical systems that can’t improvise: banking, insurance, the public sector, healthcare, justice, telecom, energy, industry and defense. Your systems can be decades old or recent: it works on the mainframe and on your APIs.

How does approval work?

You decide which steps wait for a person and who approves them. By default, Two-Key Approval applies, the four-eyes principle: anything that changes data waits for someone else, never the person who started the task, unless an admin allows it with a reason that goes into the run record. The approver sees the exact values of the change first, and if those values change, the approval no longer counts.

Where is my data? Is any model trained on it?

Inside the European Union. On the Reelease service, services run in Madrid and the encrypted database is in Frankfurt, and the AI models Reelease offers also run in the EU. Your databases are looked up on each task, not copied, and no model is trained on what the agent reads.

Which technologies does Reelease understand, and which can it act on?

It understands and documents the mainframe (COBOL, CICS, PL/I, JCL, IMS, VSAM, Db2 and MQ), Oracle (PL/SQL, Forms and Reports), IBM i (RPG, CL and DDS), Natural and Adabas, .NET Framework, Java, PowerBuilder, Delphi and VB6, Python, Ruby, NestJS, GraphQL and React. An agent acts through five paths: CICS transactions on the core; your web APIs (REST, SOAP, GraphQL and PL/SQL web); your Oracle, SQL Server and PostgreSQL databases through the Client Bridge; the operator's computer, for desktop applications; and each person's work account or the organization app, for Microsoft 365, Gmail, Jira, Confluence, Azure DevOps, LinkedIn, Slack and GitHub. What has no direct path, such as JCL or web front ends, it understands so you can document and renew it.

Does it work with decades-old systems?

Yes, and with modern ones too. Reelease understands the mainframe (COBOL, CICS and PL/I), IBM i, Natural, PL/SQL, Oracle Forms and Reports, .NET Framework, PowerBuilder, Delphi and VB6, as well as Java, Python, Ruby, Node, React and your APIs. Agents run what you publish, and there’s nothing to migrate first: you start on what you already run.

Does Reelease migrate my COBOL?

It isn’t an automatic migration. Reelease documents your COBOL, gives agents the executable use cases you publish and, if you decide to replace it, what it understands about the system (business rules, screens and flows) is your starting point. You replace it piece by piece, at your own pace, and everything else keeps running. The market calls this legacy modernization.

What does each run leave on record?

Its run record keeps the complete audit trail of the task: the Sealed Plan and its hash, who approved each change and when, every action on your systems with its outcome, and the rules that applied. You can export it and put it on legal hold, and it's kept even if you delete the agent. A complete sample run record is on the site for you to open.

Does it help with DORA, NIS2 and the EU AI Act?

It helps. The EU AI Act, DORA and NIS2 ask for human oversight, change control and traceability, and Reelease gives you controls and evidence for each, mapped by article: the run record supports the logging of Article 12 of the AI Act, and the approvals you define support the human oversight of Article 14. For DORA (Articles 28 to 30), you get the documentation to assess us as an ICT provider. The assessment is yours; we give you the evidence, the DPA and the subprocessor list.

How much does it cost?

You can start free, with one system and one agent. To try Reelease on your own systems with hands-on help there is the Pilot, and for production, Pro: both have published prices. Enterprise is custom, by contract.

See plans and pricing

Start with a process that matters to you

We’ll show you Reelease on a case like yours and, if it fits, set up a scoped pilot.

You choose the cookies

We use technical storage to remember your choice and, only if you accept, analytics cookies (Google Analytics) to understand how the site is used. You can change your choice at any time. Learn more in our cookie policy.