Security and LGPD

Control over every piece of information and every action.

You set the owner and the audience of every article, and the owner, rules and budget of every agent. And everything that people, AI and routines change is recorded.

Governed memory

You know how far to trust each piece of information.

Every Brain article shows who verified it and whether its review is up to date. Changes written by AI and contradictions found by automatic review appear in the review queue. Agent search can give more weight to what has been verified.

  • Verified

    A person verified the article, and their name and the date are recorded.

  • Approved

    Reviewed and within the review deadline set for the article.

  • Pending review

    A change written by the AI is waiting for a person's approval.

  • Stale

    Past its review deadline. The article is flagged until it is verified again.

  • Conflicting

    Automatic review found pieces of content that contradict each other.

Controls

What comes with the platform as standard.

Access, agent and data controls that your security team can check right on the screen.

Product screen, with sample data

Access profiles

Each profile sets which models, agents, skills and memory people can use.

Per-company isolation

Each customer's data is isolated. No query crosses from one company to another.

Audience per article

Each article defines whether all agents can read it or only internal assistants. Agents that talk to customers do not read internal articles.

Audit

A record of configuration changes and of who approved or rejected each agent action, plus the history of every run with its cost and each step.

Personal data masking

Email, CPF, CNPJ, phone and card numbers can be masked in logs and answers. You switch it on for each agent.

Per-tool rules

For each tool an agent can use: allow, ask for approval or block.

Human approval

Agent actions can wait for approval. You set the owner and the approvers of each agent.

Budget with an automatic stop

A monthly budget per agent, with a warning and an automatic stop when the cap is reached.

AI publishes only with approval

What the AI writes into the Brain can go through a person's review before it reaches the agents.

The memory is yours

Export of the whole wiki in the Open Knowledge Format (OKF), as Markdown files you can take anywhere.

LGPD

Built for Brazil's data protection law.

Product, contracts and support in Portuguese, designed around the principles of the LGPD.

Principles behind the product

  • Purpose and necessity: process only the data needed for the stated purpose
  • Transparency: a record of who changed what and of what each agent did
  • Security and prevention: access controls, isolation and masking
  • Accountability: an audit trail to demonstrate compliance

Data subject rights

  • Confirmation of and access to the data
  • Correction of incomplete or outdated data
  • Anonymisation, blocking or deletion
  • Portability
  • Information about sharing
  • Withdrawal of consent
Read the Privacy Policy

FAQ

Questions about security

Does Kluco depend on one AI vendor?

No. The memory stays in the Brain, and whichever AI tool the company chooses queries it over MCP, the open connection standard, or through the API. Switching vendors does not change where the memory lives or who can access it.

Can I take the memory with me?

Yes. The whole wiki can be exported in the Open Knowledge Format (OKF), as Markdown files with each article's sources and metadata.

Who decides what each person and each agent can see?

The company. Access profiles define models, agents, skills and memory per profile, and each article defines whether all agents can read it or only internal assistants, which do not talk to customers.

For your security team

Review the controls with us.

Bring your security and data protection team to a session with us and see each control in the product.