Introduction
Confluye is an AI-native workflow platform for building, running, and operating automations across your organization.
Confluye lets you design workflows on a visual canvas, run them on demand or on a schedule, and operate them with full observability. This documentation covers building workflows, the feature set, platform administration, and the REST API.
What is Confluye
Confluye is an AI-native workflow platform. You compose workflows from blocks — triggers, AI agents, logic, data, and integrations — on a visual canvas, then run them manually, on a schedule, from a webhook, or via the API. AI is a first-class citizen: agent blocks use your own model keys (BYOK), knowledge bases ground answers in your content, and the Workflow Copilot helps you build and debug. It is built for teams, with workspaces, roles, credentials, audit logs, and enterprise SSO/SCIM.
Core concepts
The unit of isolation. It owns your workflows, data, credentials, and members.
A graph of blocks that runs from a trigger to an output.
A single step — trigger, AI agent, logic, data, or integration.
One execution of a workflow, with per-node status and output.
A cron trigger that runs a deployed workflow automatically.
Indexed content that grounds AI blocks via retrieval (RAG).
Personal context that follows you across agent conversations in one workspace.
A private conversation, memory, and website login for each external customer of a product you build.
Child agents that work in parallel with reduced authority and fixed limits.
Memory, website sessions, and automations
These capabilities help agents keep context and run recurring work. Availability depends on your deployment and on your role in the workspace: memory and end user features are off until an operator enables them.
| Capability | Who it serves | Guide |
|---|---|---|
| Private agent memory | One signed-in member in one workspace | Private agent memory |
| End user memory | Each external customer of a product you build | End users and memory |
| Saved website sessions for a member | You, for your own agent and workflows | Saved website login with Codex or Claude Code |
| Website logins for end users | One customer at a time, through a login link | Website logins for end users |
| Scheduled workflows and adaptive jobs | Anyone who manages automations in a workspace | Scheduled Tasks |
A saved session is reused until the website expires it; then the person signs in again. Member memory and end user memory never mix.
How the docs are organized
- Getting Started — orientation, a first end-to-end workflow, and a quick reference.
- Building Workflows — the canvas, the block reference, running and debugging, deploying, and versions with git review.
- Features — the Command Center, Copilot, private memory, end users and their memory, bounded subagents, knowledge bases, databases and SQL, docs, scheduled tasks, logs, mailer, skills, custom tools, and MCP.
- Platform — organizations and workspaces, enterprise controls, API keys, BYOK, and keyboard shortcuts.
- API Reference — authenticating and calling the
/api/v1REST API by domain.