Transparency & Trust

AI Systems Disclosure

OmniGate uses autonomous AI agents to automate workflows on your behalf. This page explains how those agents work, what controls you have, and where the current boundaries lie.

Last updated: March 30, 2026

01 — How it works

What Autonomous AI Agents Do

OmniGate’s AI agents are software programs powered by large language models (LLMs) that can plan, reason, and take multi-step actions without a human approving every individual step. You define a goal; the agent figures out how to accomplish it.

In practice, agents can read data from connected services, draft content, trigger API calls, generate reports, and hand off tasks to other agents — all within the boundaries you configure.

🔗

API & Integration Access

Agents interact with connected tools (GitHub, Gmail, CRMs, databases) using the credentials you authorize — and only those.

🧠

LLM Reasoning

Each step is processed by an LLM (currently Anthropic Claude) that interprets context, selects the next action, and produces outputs.

🔄

Multi-Step Execution

Agents can chain dozens of actions in a single run, including calling sub-agents and branching based on intermediate results.

📋

Persistent Memory

Agents store execution logs, generated reports, and task history so they can reference past context in future runs.

02 — You’re in control

Your Governance Controls

Autonomous does not mean unaccountable. OmniGate is built with layered controls so you can see exactly what agents are doing, override decisions, and define hard limits.

⏸️

Pause & Resume

Disable any agent instantly from the dashboard. Active executions are stopped gracefully and scheduled runs are suspended until you re-enable.

📜

Full Audit Trail

Every action an agent takes is logged with timestamps, inputs, outputs, token cost, and the model version used — queryable at any time.

Approval Workflows

Configure agents to pause and request human sign-off before executing high-impact actions like sending emails, creating PRs, or publishing content.

🔒

Scoped Permissions

OAuth tokens are granted per-integration with minimum necessary scopes. Agents cannot access services you haven’t explicitly connected.

💰

Cost & Usage Limits

Set per-agent and per-company operation budgets. Agents that approach a limit pause and notify you before incurring additional cost.

🗑️

Data Deletion

Delete any agent, execution record, or stored credential at any time. Deletion is permanent and cascades through all linked records.

03 — Current boundaries

Current Limitations

We believe in being upfront about what OmniGate’s AI agents cannot or should not do. These limitations reflect both technical constraints and deliberate policy decisions.

04 — Get in touch

Contact Us with AI Concerns

If you observe unexpected, harmful, or concerning agent behaviour — or have questions about how OmniGate’s AI systems handle your data — please reach out directly. We respond within one business day.

✉️

AI Safety & Concerns

Email us at omnigate@polsia.app with the subject line AI Concern. Include your account email, a description of the agent behaviour, and — if possible — the execution ID from your audit log.