(Run Everything. Verify Everything. Record Everything.)
by Torvexis LLC
Self-hosted automation built for data sovereignty, full observability, and compliance. Credentials are encrypted at rest, access is role controlled, token spend is tracked and budgeted, and every execution is audit ready. Drop your code into the module wrapper, add it to the folder, and it loads without a rebuild.
Modules do the work. Workflows control the flow. You control the data.
No SDK. No package manager. No publishing pipeline. A module is a Python script and a config file in a folder. Write your logic, match the standard wrapper, and drop it in. Hit refresh and the platform scans the module folder and picks it up. No rebuild. No redeploy. Each module runs in its own isolated subprocess, so a failing module never crashes the system or breaks other workflows.
Zero coding experience needed. The prebuilt wrappers give AI coding tools enough structure to generate a working module from a plain description. Drop it in, wire it into a workflow, and check the execution trace and diagnostics to see exactly what went in and what came out. Developers get full control. Non-developers get full access.
Every connection between modules flows as JSON. A module only receives data from nodes you explicitly wire into it. Nothing else. If a workflow has five steps, the third one only sees output from whatever you connected to it. No ambient data. No silent forwarding.
This is where data exposure and token cost stack up in most tools. An LLM module that only needs a summary field should never receive your entire dataset. Route the field, not the payload. Less data in means less cost, less exposure, and a smaller attack surface.
No black box. Every workflow run is traced node by node with full input data, output data, duration, and errors captured at each step. Attach a diagnostics tap to any connection to inspect live data mid execution. When something fails, you see exactly which node, exactly what it received, and exactly what went wrong. No digging through cryptic logs across multiple systems. No silent failures. No guessing whether something actually ran.
Workflows can be triggered from anything you wire up, but the platform ships with a chat interface out of the box. Link to a workflow, type a message, and results stream back in real time over WebSocket. No dashboard navigation, no external tool required. Good for testing, good for quick runs. Every chat execution gets the same full trace and logging as any other trigger method.
Build a sequence of steps once and call it from any workflow. No duplicated logic. No fixing the same bug in five places. Tasks have their own Input and Return nodes, full version history, and dependency tracking. Disable a task and every workflow that references it is automatically paused. You always know what uses what, and a change in one place applies everywhere.
Every running service registers itself and reports its state. Workers, WebSocket connections, and input modules all show up on a live status page with current health, poll timing, and processing counts. No green dashboards hiding silent failures. The sidebar carries a persistent health indicator so you know the moment something drops without navigating anywhere. Dedicated metrics endpoints track request performance, execution stats, and token cost across the whole platform.
Drag modules onto a canvas and connect them visually. No endless scrolling through linear step lists that fall apart after six actions. Conditional branching, parallel execution, loops, gates, and merge nodes are all built in. Connection pins on each node can be dragged to any edge so you arrange your canvas however makes sense for your workflow, not however the tool decided it should flow. Every save creates a new version you can roll back to.
Users, roles, credentials, system settings, session policies, and account lockout rules are all managed from a single centralized dashboard. Granular role based access controls who can build, who can run, and who can view. Credentials are stored in an AES-256-GCM encrypted vault with per credential salt, decrypted only at execution time, never logged. Full activity tracking shows who changed what and when. White glove branding that goes beyond a logo upload. Page backgrounds, sidebar colors, button colors, and per module category colors are all configurable to match your environment. Session timeout, failed login lockout, and security policies are adjustable in system settings. Roles, permissions, version history, scheduling, execution logging, encrypted credentials, and full theming are all fully included out of the box.
Whether you write code or not, if you care about where your data goes and what your automation is doing, this was built for you.
If you write code, you have full control. If you don't, the module wrapper and AI coding tools get you there. Same platform, same access.
Your infrastructure, your credentials, your data. Nothing leaves your network unless you built it to.
Role based access, per user token budgets, and workflow permissions scale from a solo operator to a full department. Same platform, no tier gating.
Full execution trace, encrypted credentials, role based access, activity logging, and configurable data retention. Audit ready out of the box.
Everything ships. No upgrades required.
Full audit trail for every workflow run. Input, output, duration, and errors captured at every node.
Every save creates a new version. Roll back to any previous state. Full history preserved.
AES-256-GCM encrypted storage with per credential salt. Decrypted only at execution time, never logged.
Owner, Manager, Developer, User. Granular permissions per workflow and feature.
Cron based scheduling with automatic startup. Set it and walk away.
Monitor LLM token usage per user, per workflow, per model. Set daily limits.
Live status page monitoring backend, WebSocket, and module connections.
Cancel a running workflow mid execution. No waiting for it to finish or fail on its own.
Share workflows with specific users or keep them private. Owner controls visibility.
Configure alerts for failures, errors, and token limits. Email, webhook, or in app.
Configurable cleanup for execution logs, chat history, and activity records.
Detects an empty database and walks you through creating the first admin user.
Configurable session timeout with activity tracking on every authenticated request.
Failed login lockout with configurable attempt limits and per IP rate limiting.
Condition, switch, merge, loop, delay, gate, and transform. All built into the execution engine.
In active development. More details coming with the release.
The agentic engine is a separate execution mode designed for larger, multi-step projects where a standard workflow is not enough. It handles planning, task ordering, token budgeting, and iterative execution with the goal of getting more done with fewer tokens.
This section will be expanded when the engine is ready for public use.
Your infrastructure. Your rules. Standard stack with zero vendor lock in.
Open source release coming soon. Free for personal and non commercial use. Commercial licensing available on request. Interested?
Email: revere@torvexis.com