# Open Docy

> Open Docy (also written OpenDocy) is the free, open-source and self-hosted edition of the Docy AI Agent OS. It is a visual builder and runtime for AI agents: you compose an agent from nodes on a canvas, run it as tasks on your own infrastructure with Docker Compose, and keep a run log and audit log of what each agent decided and why. A person can be asked to decide at any step.

Open Docy is maintained by Docy AI (https://docyai.com). Website: https://opendocy.com. Source: https://github.com/OpenDocy/opendocy.

## What it does

- Visual agent canvas: drag in nodes, connect them, debug one node at a time, save drafts and publish.
- Task runtime: seven task states (Not Started, Queued, In Progress, Action Required, Pause, Completed, Failed), a run log and flowchart for every run, results and downloads.
- Human in the loop: the Interaction node pauses a run at Action Required and asks a person for text, a choice or a file, then resumes.
- Rules that explain themselves: audit nodes check values against rules (equals, contains, greater than, less than, before, after) and show the value found, the value expected and the reason.
- Document templates: annotate a DOCX or PDF template; the template node fills it from upstream output.
- Audit log: who did what to which agent or task, with CSV export.

## Node types available in Open Docy

input_file, input_file.repeatable, data_extractor, data_extractor.batch, extract_signature, audit.global, audit.field, conditional, interaction, document_template, web_search, content_generator.

Knowledge, compliance_policy and custom_component nodes are visible but locked in Open Docy; they are part of Docy AI.

## Install

Requires Docker with Compose v2 (Linux, macOS, Windows with WSL2).

    git clone https://github.com/OpenDocy/opendocy.git
    cd opendocy
    docker compose up -d

Then open http://localhost:3000.

## Open Docy and Docy AI

Open Docy covers building and running agents in a personal space. Docy AI, the commercial edition, runs in the cloud or on-premise and adds building agents by chat, sealed decision records and decision replay, team spaces and roles, REST API keys, OAuth and webhooks, knowledge bases, versioned compliance policies, custom components, a pre-built agent library and usage controls. Both editions use the same node types, so agents carry over.

The app, canvas and task runtime are open source. The extraction and AI logic inside each node ships compiled.

## Typical uses

Rebate and compliance checks (energy), credit assessment (finance), invoice and purchase-order matching (accounting), due diligence and lease abstracts (property), contract and supplier review (procurement), briefings from web sources (research).

## Links

- Website: https://opendocy.com
- Source and issues: https://github.com/OpenDocy/opendocy
- Discussions: https://github.com/OpenDocy/opendocy/discussions
- Security reports: https://github.com/OpenDocy/opendocy/security/advisories/new
- Docy AI: https://docyai.com