Skip to content
DuckBug
Menu

Manage projects.Put AI to work.

DuckBug brings tasks, knowledge and monitoring into one workspace. Your team and AI do the work; you see the progress and results.

For development teams and solo developers, legal professionals, managers, sales and support teams.

ProjectTeam and materials
WorkPeople and AI agents
OversightTasks and results

Tasks for people and AI.

Projects hold tasks and materials. The owner can be a colleague or an AI agent. The board shows what is in progress.

Take a closer look
AtlasProject tasks
Interactive example
To do2
In progress1
In review1
DB-204

Add CSV report export

A customer needs the report in Excel. Today they copy the data by hand.

Brief · design · repository
Goal → work → resultDuckBug

Example development task. A person can be assigned in the same way.

Stop explaining the project from scratch.

The assistant answers questions using project materials. Knowledge, meeting decisions and saved rules give it context for the next assignment.

One document for the team and agent.

Use the knowledge-base requirements in the next assignment.

Atlas · knowledge baseExample

“Which discounts can we offer?”

Sales policy: a manager can offer up to 10%. A larger discount needs approval from the sales lead.

Answer from the document

“You can offer up to 10% yourself. For 15%, get approval first.”

Discount policy · section 2
Atlas · customer meetingExample

“What did we agree at the meeting?”

The customer is piloting with 20 employees. They need access and a setup guide by Friday.

Decisions from the meeting

“Pilot: 20 employees. By Friday: provide access and send the setup guide.”

Recording analysis requires setup and participant consent
Atlas · saved decisionsExample

“How should new APIs be designed?”

Project conventions: UTC dates, a shared error format and cursor pagination.

Context for the next assignment

The assistant uses these rules when drafting requirements for a new API.

Project memory · API conventions

Error detected. AI prepares the fix.

DuckBug collects errors, logs and signals. A configured assignment finds a new issue, then agents prepare and review the fix.

Interval: 5 minutes

“Check for new errors every 5 minutes. Create tasks and take fixes through to a reviewed PR.”

01Signal02Fix03AI review04Ready
Grafana → DuckBugExample

A customer cannot place an order.

checkout-api · HTTP 5xx
TypeError · checkout.ts:42

The signal and error stack are ready to investigate

DB-128 · AI Coder

The agent fixes the cause.

The assistant created DB-128 with the error stack and assigned the agent.

− return cart.items.map(toLine);+ return (cart.items ?? []).map(toLine);
Empty-cart check passed
AI Reviewer ↔ AI Fixer

Feedback goes to the agent.

AI Reviewer

Add a test for items = null.

AI Fixer

Test added. Checks passed.

Follow-up AI review passed
Chain completed
PR #42

A reviewed fix, ready to merge.

ChecksAI reviewDB-128

The error, code and checks are linked to the task. In this example, a developer approves the merge.

Example of a configured automation.

Where is work getting stuck?

Organization Pulse shows the owner what needs attention across projects: blockers, pending reviews and failures. Open the cause from the summary and assign the follow-up.

  • See which projects need help.
  • Open the blocker, error or agent run.
  • Assign the investigation to a person or agent.
2Awaiting reviewNeeds attention
0UnassignedAll clear
0Roadmap risksAll clear
1BlockersNeeds attention
0Agent failuresAll clear
0InfrastructureAll clear
3ErrorsNeeds attention
0LogsAll clear
Atlas and CRM · two PRs await review

Report export and customer search are ready for review. Changes and owners are accessible from the summary.

Tasks have owners

This example has no tasks without an owner.

No identified roadmap risks

Review roadmap health across projects.

Website · launch awaits design approval

A task is blocked. The summary shows what delays the launch and who needs a response.

No failed agent runs

A failed agent run appears in the overview.

No active signals

Issues from connected infrastructure appear here.

API · three error groups

Errors are grouped. Open the events and ask the assistant to investigate.

Logs need no attention

Track problematic log groups alongside the rest of the work.

Sample overview. Choose a direction.

Delegate more than code to AI.

A development plan, contract comparison, meeting follow-up or sales proposal. Describe the outcome and provide the materials.

Choose a workflow
Your assignment

“Break the move to API /v2 into tasks.”

Current API description and migration goal

Before DuckBug

Turn requirements into a plan by hand.

Development planExample

Move to API /v2

  1. Define the new contractFields, errors and /v1 compatibility
  2. Implement /v2 handlersSeparate tasks for API methods
  3. Prepare client migrationChecks and an upgrade guide

The assistant creates tasks when permitted.

Your assignment

“Turn the meeting notes into team tasks.”

Meeting notes with decisions and owners

Before DuckBug

Copy decisions out of meeting notes.

Meeting follow-up tasksExample

New website launch

  1. Approve the homepage designMaria · by Wednesday
  2. Prepare section copyAnton · by Thursday
  3. Test the enquiry formIlya · by Friday

Tasks use the owners from the notes, when access is permitted.

Your assignment

“Draft a proposal from the customer’s request.”

Customer request, service description and price list

Before DuckBug

Write the same kind of email from scratch.

Proposal draftExample

Support rollout for 20 employees

  1. ScopeWorkspace setup and team training
  2. Questions to clarifyStart date and customer contact
  3. AttachmentsService description and terms from the price list

The manager reviews the proposal before sending.

Stop answering the same question.

AI answers customers using your knowledge base. When the team is needed, turn the conversation into a task with its history attached.

  1. 01
    Choose articles for answers

    You select the articles the assistant can use.

  2. 02
    Create a task from the conversation

    Link the conversation to a team task.

  3. 03
    Check open conversations

    History, status and analytics in the support app.

Atlas supportExample

My team invitation has not arrived. What should I do?

AI assistant

Check the email address and spam folder. An administrator can resend the invitation.

Article: Inviting teammates
Answer based on the team’s instructions.

Example of a configured widget with published knowledge.

Atlas supportExample
AI assistantArticle: Inviting teammates

The invitation still has not arrived after resending.

SUP-12 · OpenInvestigate invitation delivery

The task is linked to the conversation. History stays available to the team.

Work the way you know. Without the chaos.

Keep your IDEs, repositories and AI. DuckBug connects them to tasks and results, so you can start with your current workflow.

  • OpenAI

    ChatGPT · Codex

  • Claude

    Claude Code

  • Gemini

    Gemini CLI

  • Cursor

    Cursor Agent

  • Kimi

    Kimi Code

  • OpenCode

    Multi-model

Choose a model for the work

Codex, Claude, Gemini and other agents. More models are available to OpenCode through OpenRouter.

Work from your usual assistant

Connect DuckBug to any MCP-compatible AI assistant and work with tasks and knowledge from there.

Connect the tools you already use.

Integrations give DuckBug access to code, events and tools. Choose the connections each project needs.

GitHub
GitLab
Gitea
DuckBugRepositories and pull requests
Code stays in your Git host

Repository connected to the project

An agent can work on the code. Changes and pull requests are linked to tasks.

Code stays in your repository. Connect your Git account and choose a project.

Prometheus
Grafana
Argo CD
DuckBugSignals and context
Events arrive in DuckBug

Monitoring sends signals

Infrastructure alerts are available in the project alongside application errors and logs.

Webhooks for Prometheus, Grafana and Argo CD. Other systems require a configured adapter.

MCP
SSH
SQL
DuckBugAuthorized tools
Actions in your systems

The assistant gets tools

MCP connects external services. SSH provides server access; SQL helps read permitted data.

Available actions depend on connected tools, credentials and user permissions.

Keep working from any device.

Tasks, materials and discussions are available in DuckBug’s web, desktop and mobile apps.

WebDesktopMobile

App download links have not been published yet.

DuckBugAtlas
In review
DOC-17

Approve the pilot terms

Pilot terms.pdf

With the team for review
DuckBug
DOC-17

Approve the pilot terms

Dates agreed. I left a comment on the payment section.

Discussion and materials

An illustrative task across devices.

Need your own servers? Self-host DuckBug.

Use DuckBug in the cloud or deploy it on company infrastructure.

We operate the cloud servers.

With self-hosting, your team manages hosting and updates.

DuckBug cloud

DuckBugProject data
Hosted on
The service’s cloud
Operated by
The service team
Start in the cloud

Your company infrastructure

DuckBugProject data
Hosted on
Your servers
Operated by
Your team

Self-hosted deployment. AI providers and external services are connected separately.

AI does the work. You set the boundaries.

You choose the repository, tools and agent permissions. Its actions and results can be reviewed.

PermissionsWithin authorized projects and tools
ReviewAction history and results stay with the task
LimitsConditions and boundaries for recurring work
How is an assistant different from an agent?

You talk to the assistant: ask questions, give assignments or request tasks. An AI agent does assigned work, such as changing code or reviewing a PR. Results are available in the project.

Does AI make every decision itself?

You define the task, connected tools, permissions and limits. Results and actions are available for review. Execution policy depends on the configured agent and assignment.

Do we have to move our whole process?

Start with one project and keep your IDEs, Git and AI. Connect the integrations you need. MCP lets compatible external assistants work with tasks and knowledge.

Can I start without a team or infrastructure?

Yes. A solo developer can create a personal project and a first task. The cloud version requires no server deployment. Add colleagues and more tools later.

Start with work that is already waiting.

Create a project, add one task and its materials. Connect other tools when you need them.

  1. 01

    Create a project

    For yourself or your team.

  2. 02

    Add a task

    Goal, materials and expected result.

  3. 03

    Choose an owner

    A teammate or connected AI agent.

Get started