[KongchangAI]
Product

Agentboard

An open-source AI agent task collaboration framework featuring a kanban-style card management interface. It supports asynchronous collaboration between humans and AI agents, with a built-in five-state transition mechanism and a context management system.

Timeline (last 90 days)

Oct 6

Agentboard is an open-source agent task collaboration framework whose interface is in kanban form, with AI agents proactively taking over cards and autonomously completing tasks at the appropriate time in the background.

Unverified50%
Oct 6

Agentboard was shared and built by a developer with 17 years of experience on Bilibili.

Unverified50%
Oct 6

Agentboard's kanban is designed with five status columns: Inbox, Ready, In Progress, Needs Input, Review, and Done.

Unverified50%
Oct 6

In Agentboard, each card has a timestamped timeline at the bottom that records messages between humans and agents, status changes, and executed operations, to ensure that the agent always has the correct context.

Unverified50%
Oct 6

When an agent starts up, it scans the Ready column, autonomously determines which tasks need attention, and then picks up cards and moves them to the In Progress state.

Unverified50%
Oct 6

Agentboard's context system consists of a set of plain Markdown files that store contacts, information, connections, and other data accessible to agents, which agents can read and write.

Unverified50%
Oct 6

Agentboard has a dedicated area for storing keys; agents cannot read the plaintext of keys but can still use them in their work.

Unverified50%
Oct 6

Ops Cards are distinct from task cards and are used to represent things the system itself needs, such as changes to access permissions, decisions, context, or connections; when an agent encounters an obstacle, it can create an Ops Card on its own to block the current work until a human provides additional information.

Unverified50%
Oct 6

Agentboard is connected to the GitHub repository holding its own code, and the author uses the tool to build the tool itself, creating an issue for every modification and converting it into a new version.

Unverified50%
Oct 6

In the demo case, the agent fixed a missing CSS issue for empty card prompts on the desktop, added one line of code based on the existing mobile styles, and submitted a pull request on GitHub.

Unverified50%

All Facts (10)

Unverified

Agentboard is an open-source agent task collaboration framework whose interface is in kanban form, with AI agents proactively taking over cards and autonomously completing tasks at the appropriate time in the background.

50%
Unverified

Agentboard was shared and built by a developer with 17 years of experience on Bilibili.

50%
Unverified

Agentboard's kanban is designed with five status columns: Inbox, Ready, In Progress, Needs Input, Review, and Done.

50%
Unverified

In Agentboard, each card has a timestamped timeline at the bottom that records messages between humans and agents, status changes, and executed operations, to ensure that the agent always has the correct context.

50%
Unverified

When an agent starts up, it scans the Ready column, autonomously determines which tasks need attention, and then picks up cards and moves them to the In Progress state.

50%
Unverified

Agentboard's context system consists of a set of plain Markdown files that store contacts, information, connections, and other data accessible to agents, which agents can read and write.

50%
Unverified

Agentboard has a dedicated area for storing keys; agents cannot read the plaintext of keys but can still use them in their work.

50%
Unverified

Ops Cards are distinct from task cards and are used to represent things the system itself needs, such as changes to access permissions, decisions, context, or connections; when an agent encounters an obstacle, it can create an Ops Card on its own to block the current work until a human provides additional information.

50%
Unverified

Agentboard is connected to the GitHub repository holding its own code, and the author uses the tool to build the tool itself, creating an issue for every modification and converting it into a new version.

50%
Unverified

In the demo case, the agent fixed a missing CSS issue for empty card prompts on the desktop, added one line of code based on the existing mobile styles, and submitted a pull request on GitHub.

50%

Source Articles