Without a shared workflow
Work jumps between prompts, pull requests, documents, and status meetings. It is unclear what the agent saw, what changed, and who has reviewed it.
Open-source agentic Kanban
Move software forward.
Keep people in control.
KanGentiX brings AI agents into the delivery workflow your team already trusts: tickets, shared context, reviews, approvals, and pull requests—on one self-hosted board.
Human-gatedPeople decide when every ticket moves forward.
Bounded executionAgents work in disposable, limited sandboxes.
Open sourceSelf-host it and shape the workflow around your team.
The problem
AI moves quickly.
Delivery needs care.
Agent output is easy to generate and hard to govern. Context gets scattered, decisions disappear into chat, and a finished run can be mistaken for finished work. KanGentiX gives agent-assisted delivery a clear, shared path with your team at the controls.
With KanGentiX
Each ticket carries its context, generated documents, discussions, run history, approval state, and pull requests from backlog to done.
How it works
One ticket.
A deliberate path.
Design the stages that fit your team. Combine human work, agent-generated documents, and implementation steps. Agents never advance work on their own.
Shape the ticket
Start on the shared backlog or board with the context your team needs.
Set the workflow
Configure manual and agent steps, reviewers, and approval requirements.
Run an agent
Generate a document, answer questions, or prepare repository changes.
Review together
Discuss results, inspect redacted logs, and approve the current work.
Advance explicitly
People move the ticket forward and merge pull requests separately.
Made for accountable delivery
Useful agents.
Visible boundaries.
KanGentiX is built around the details that let a team use agents with confidence, rather than hoping a black-box automation did the right thing.
✓
Configurable approvals
Choose eligible reviewers and approval counts per step. New work invalidates stale approvals, so people review what actually changed.
⌘
Sandboxed agent runs
Run agents in a fresh Docker or E2B sandbox with time and resource limits. GitHub credentials stay in the trusted app.
↗
Verified pull requests
Agents prepare changes; KanGentiX validates and publishes the resulting pull request through GitHub's API.
▤
Shared working context
Keep ticket details, Markdown documents, attachments, conversations, activity, and execution history together.
◌
Project-level control
Set membership, permissions, workflow rules, model providers, sandboxes, and concurrency limits for each project.
⌕
Traceable by default
Inspect run status and history at the ticket level, with secrets redacted before logs are retained or displayed.
Your workflow, your infrastructure
Open source and built to be self-hosted.
KanGentiX pairs a React interface and Express API with PostgreSQL, and supports disposable Docker or E2B environments for agent work. Deploy it where your team needs it, connect the providers you choose, and retain control of your projects and credentials.
Read the README ↗# Start KanGentiX locally
$ cp .env.example .env
$ docker compose up --build -d
# Then open
→ http://localhost:3000
$ cp .env.example .env
$ docker compose up --build -d
# Then open
→ http://localhost:3000
Ready to explore?
Put AI delivery on a board your team can trust.
Read the source, run it locally, and build a workflow that keeps speed and accountability together.