Your account
Sign in to enter the hosted workspace. Access depends on the deployment's admission configuration and your account's entitlement.
Your implementation will change. Keep the conversations, decisions, and lessons that help you understand what to build next.
Burdgen is Laravel first, built for developers who want to understand their projects as well as change them. Ask about your architecture, explore its patterns and tools, and find the concepts you need to learn. The aim is to grow your understanding alongside the code.
Import Claude Code history and browse sessions by project. Read the messages in order, including recorded tool activity and the answers you gave. Search decisions by their text, or narrow them to a branch, file, session, project, or date range.
When a choice deserves a clearer explanation, give it a curated decision record. Add a title, rationale, related files, and citations. You can write one yourself or promote a decision from imported history.
The history area also has readers for imported plans, memory documents, configurations, artifacts, pastes, and terminal commands. Availability depends on the sources you import. These readers do not imply that every source has been synchronized to a hosted account.
Burdgen separates its record of a request from the provider that executes it. A run keeps the selected target, its reported capabilities, and its execution identity so you can tell which environment did the work.
AI features such as summaries and commit messages have separate provider and model settings. Local model configuration and hosted provider configuration depend on the feature. The coding runtime has its own model options and authentication.
Local models are an important part of where Burdgen is headed. We want you to be able to work with models on your own hardware and choose a development environment that fits the project. That support is still being developed.
Model choice and environment choice should not force you to rebuild the history of your project. Burdgen keeps the request, decisions, and results in its own records, while available tools depend on the connections you configure.
Burdgen connects a project to its repositories and recorded observations. The reader gives you entry points into files, symbols, references, pull requests, tests, and supported architecture and database inspections.
Those records have limits. An unindexed repository is not an inspected project, and last week's scan is not a live view of today's checkout. Burdgen labels absent, partial, and stale evidence so you can judge what you are reading.
Pull request reading keeps commits, file changes, and review discussion accessible in the same workflow. When generated explanations are available, they are interpretations of the source material. The underlying records remain the place to check a claim.
Your preference profile describes how you like to build across projects. It can include architectural principles, testing expectations, tool preferences, and the tradeoffs you are willing to make.
The manifesto is the collection of opinions you choose to put into practice. Give each one a rationale and a scope. A preference for thin Laravel controllers may be broadly useful; a particular queue configuration may belong to one application.
A user-visible profile makes tradeoffs atomic: each choice has its own rationale, benefits, costs, and scope. You can inspect and customize that choice independently. For example, you might usually favor fewer dependencies but accept a maintained package for a particular capability. That exception should not change your preferences everywhere else.
We are developing a way to configure that manifesto for each project, carrying the relevant opinions into instructions, plans, and review. You decide what becomes guidance. A passing comment or a one-off compromise should not silently become a permanent rule.
Deliberation gives you a place to challenge those opinions. Keep the alternatives and evidence, record an exception when it is justified, and revise your preferences when experience changes your mind.
Versioned project instructions already preserve what a task received. The broader profile and manifesto workflow builds toward applying your judgment consistently without treating every project as identical.
Read how Deliberation shapes your manifestoAn execution request records the instructions you submitted, the attached inputs, the selected target, and the permissions the job needs. Burdgen stores the request before dispatching the work and gives it a permanent page.
Task plans show dependencies, readiness, and available actions. Run records show the execution events, output, result, and artifacts the runner reports. You can leave the page and return later without making your browser responsible for keeping the agent alive.
Corrections create linked requests. Re-planning keeps the earlier plan. These are useful distinctions when you need to understand why a second attempt differed from the first.
Each grants a different kind of access. An account alone does not authorize a computer to run code.
Sign in to enter the hosted workspace. Access depends on the deployment's admission configuration and your account's entitlement.
Connect GitHub, choose an available repository, and confirm it before Burdgen starts ingestion. The cloud project belongs to your account.
Use a one-time pairing code to connect a Mac or Linux machine to a repository and local workspace. Its credential authorizes runner communication separately from your browser session.
Choose the target and declare permissions for the work. Target checks consider the repository, workspace, reported capabilities, and authorization before dispatch.
Burdgen is in private preview. The source setup is available to people with repository access; there is no published desktop installer.
With a paired local runner, the agent runs on your Mac or Linux machine in its authorized workspace. The hosted application records and dispatches the request. It does not execute that local job inside the web request.
Requests and their uploaded inputs, connected repository records, runner events, output, results, and uploaded artifacts. Those records can include source code and other project content. External model or tool use also depends on the execution runtime you configure.
Yes, for local execution. The runner must be able to poll the hosted service and run its configured tools. Closing the browser is different from putting the execution machine to sleep.
Yes. The runner enrollment settings offer a confirmed disconnect and revoke action. Revocation blocks new work without erasing run history or deleting files on the computer. The runner must pair again before it can take new work.
No. Completion describes the execution. Read the result and verification evidence, inspect the change, and decide whether it meets the intended outcome.
The current conversation and decision importer reads Claude Code history. Other execution integrations can produce run records, but that does not give them the same conversation import support.
No. Hosted project reading uses repository records and available inspection evidence. A connected execution target is needed when you want to dispatch coding work.
Leave your email on the interest list. We will reach out when more preview places are available.