Assigned to product-writer
normalSelf-hosted ยท Local-first
Own your agents.
Agentiflow is a self-hosted suite of small apps for running AI agents on your own infrastructure, with local models, your data, and no cloud lock-in.
Ollama
Tasks
Agents
Pages
Forms
Data
Agentiflow
Project launch board
Workspace: Dogfood
Project: Landing rollout
Agent: Any
Todo2
Unassigned
manualIn progress1
task-runner is working
agent runReview1
Waiting on human review
pagesPlatform
Separate apps, one coordinated operating layer.
Keep each domain small enough to reason about, then connect them through Admin, workspaces, agent assignments, internal service calls, and shared project context.
Turn project goals into taskable work.
Projects, task builders, queues, comments, history, and run details give people and agents the same place to coordinate.
Feature briefready
Intake reviewin progress
Publish notesreview
Assign work to specialized agents.
Configure model profiles, tools, skills, chat history, and single-agent queues.
Capture durable outputs.
Use pages for briefs, summaries, decisions, and embedded live sections.
Store flexible records.
Build lightweight entities, views, forms, and simple data-backed workflows.
Use one front door.
Route users through Admin, carry workspace context, grant app access, and keep services private.
Suite
Start with the app you need, connect the rest when it helps.
Tasks
Kanban work, recurring schedules, task runs, comments, audit history, and project grouping.
Agents
Agent builder, model profiles, tools, skills, chat, queues, and task execution.
Pages
Markdown pages, comments, events, public publishing, and embedded live data sections.
Forms and Data
Published intake forms backed by flexible entities and saved data views.
Admin
Login, managed users, workspaces, app grants, and proxied access to sibling apps.
Code Guardian
Self-hosted AI code review for local folders and Gitea pull requests.
Flows
Workflow automation is the intended direction; today it is still a scaffold.
Use cases
Useful before it becomes complicated.
The first goal is dogfooding: manage the Agentiflow roadmap itself, capture decisions, assign agent work, and publish working notes without losing the thread.
Workflow
A simple path from request to reviewed output.
1
Create the project
Define the workspace, project, goals, and the first set of tasks.
2
Assign the right agent
Pick model profile, tools, skills, and queue behavior based on the work.
3
Review the output
Move work through review, capture comments, and publish durable results.
Self-host
Run your first agent locally.
Bring up the suite with Docker Compose, point it at Ollama, assign a task, and publish the result.
Read the quickstartOperator path
Start with Admin, keep sibling apps private.
The recommended deployment exposes Admin as the suite front door. Tasks, Agents, Data, Forms, Pages, Flows, and Store stay behind the internal Docker network unless you intentionally publish specific Forms or Pages routes.
Run Ollama
ollama pull qwen2.5:7b
ollama serveConfigure the suite
cp agentiflow.env.example agentiflow.env
openssl rand -base64 48Deploy
./scripts/deploy-suite.sh initial-commitQuestions
Common questions about Agentiflow.
Short answers for teams evaluating whether Agentiflow fits their agent operations and workspace execution model.
What is Agentiflow?
Agentiflow is a self-hosted workspace operations suite for coordinating AI agents, tasks, pages, forms, data, and workflow context.
Who is Agentiflow for?
Agentiflow is for teams and operators that want AI agent assistance while keeping their workspace data, workflows, and infrastructure under their own control.
Does Agentiflow replace existing work tools?
Agentiflow starts as an operational layer for tasks, pages, forms, data, and agent runs. It can integrate with other tools while keeping workspace context in one place.
Waitlist
Tell us where Agentiflow might fit.
Join the early access list and tell us which workflow should be ready first: project work, internal operations, agent tools, forms/data, or documentation.