Why we built cRUD Collab

cRUD Collab TeamcRUD Collab

7 min read

We built cRUD Collab around a problem that calls and chat do not solve. The people and AI tools working on the same Product or Project rarely begin with the same trusted context, and the organisation often loses what the team learned after the session ends.

TL;DR

  • Engineering context is fragmented across people, repositories, issues, documents, calls, local environments and private AI sessions.
  • cRUD Collab connects authorised participants to shared Product and Project Context, purpose-specific evidence and governed decisions.
  • It is a browser-first collaboration and control surface, not a browser IDE or replacement for existing engineering tools.
  • AI may assist and propose, but only an authorised human can approve organisational knowledge.

The repository does not hold the whole story

A repository can show what changed. It does not always explain why the change was necessary, which alternatives were rejected, which requirement controlled the decision, which test changed the team's understanding or who approved the final boundary.

The missing context sits across issue trackers, documents, messages, calls, local notes and the memories of individual contributors. AI assistants add another source of useful but private and uneven context. Different people may ask different tools the same question and receive different answers without a shared organisational baseline.

When the team returns to the work, it spends time reconstructing what it already learned. Onboarding slows down, review repeats old debates and a later change can accidentally reverse a decision whose reason was never preserved.

A call transmits the moment

A meeting product can carry voice, video and screen pixels. Those are important collaboration capabilities, but they do not automatically assemble the relevant context before the work begins or preserve an approved outcome afterwards.

A useful engineering collaboration environment needs to connect requirements, code, issues, tests, documents, human statements and selected AI contributions to the decision they inform. It needs to show what is evidence, what is still being explored and what the organisation has approved as current knowledge.

If the room produces no prepared context, evidence relationship, governed decision or reusable memory, the team still has a meeting and the underlying problem remains.

Shared context before shared conversation

cRUD Collab organises work from the Organisation into Products and Projects. Product Context carries the purpose, terminology, requirements, architecture principles, guardrails and approved decisions that apply broadly. Project Context adds the objective, scope, repositories, issues, dependencies, risks, actions and outcomes for a defined delivery effort.

Authorised participants inherit the context that applies to their scope. They do not receive every piece of information the organisation holds. Context is assembled for the purpose, permissions and task, including the exact evidence and unresolved questions that matter.

That preparation gives people and approved AI tools a common baseline before they begin working together.

Evidence, working context and approved context

Not every document, message or model output should become organisational truth. cRUD Collab separates three trust levels so the team can explore openly without losing the boundary between observation and approval.

Evidence

What was observed, supplied or produced, with its source, identity, time, permission and lifecycle.

Working Context

Drafts, hypotheses, possible explanations and proposals that remain clearly labelled and under review.

Approved Context

The current trusted organisational position, promoted only by an authorised human Context Approver.

A challenge does not silently overwrite approved knowledge. It creates a proposed version linked to the current position. The history remains visible until an authorised approver resolves the conflict.

Rooms persist. Sessions remain specific.

A Room has an enduring purpose. It can retain its permitted context, resources, membership, decisions and history. Each live occurrence is a separate Session with its own attendance, consent, capture state, evidence and outcome review.

This distinction supports recurring work without pretending every occurrence is the same. A Daily Sync Room can continue across many Sessions. A debugging Session can end while the Room preserves the relevant decision and action. An Instant Session can begin as one-time work and become a persistent Custom Room without losing authorised history.

Consent remains Session-specific. A participant can see the current capture, retention, source and AI-processing policy before joining. Consent to a recording does not automatically publish everything captured into Shared Context.

AI participates without becoming the authority

cRUD Collab does not require a proprietary cRUD model. Organisations can approve the AI providers and tools that fit their policy. A participant can supply a permission-scoped Context Package, and selected AI outputs can return as evidence or working context.

The AI receives no more authority than the initiating person. It may summarise, extract, compare or propose. It cannot approve organisational knowledge. The provider, model, user, supplied context and output identity remain traceable where available.

This model lets teams use the tools they already value without pretending that private model history is shared organisational memory.

Browser-first, not a browser IDE

The browser is the collaboration and control surface. People can use shared context, live media, screen sharing, boards, chat, agendas, decisions and actions while continuing to work in their preferred editor, terminal, repository and issue system.

GitHub, Jira, Slack, documents, approved links and other evidence sources remain connected systems. cRUD Collab preserves the relationship between their evidence and the collaborative outcome. It does not claim to replace VS Code, Cursor, JetBrains, Git hosting, CI or the customer's development environment.

The product is stronger when it connects the tools that hold evidence instead of rebuilding each one badly.

Organisational memory is the durable value

The live Room helps the team solve a problem together. The longer-term value is that an approved outcome remains searchable, permission-aware and connected to its evidence. A later person can see what the organisation trusts, what has been superseded and which questions remain open.

Context Packages can prepare onboarding, architecture review, debugging, incident response, code review or an AI task from exact source versions. The next session begins with a better baseline. The organisation does not have to rely on the continued presence of one person or one tool to remember why the work looks the way it does.

That continuity is why we built cRUD Collab.

The product boundary in one working example

A product team is investigating whether a recent change caused a customer-facing failure. The repository owns the code and commits. Continuous integration owns its test execution. The issue tracker owns the delivery record. cRUD Collab does not replace those systems. It connects the authorised evidence they produce to the shared question, the team's working explanations and the final human-approved conclusion.

During the investigation, an approved AI tool may receive a permission-scoped Context Package and prepare a comparison of possible causes. Its response returns as a traceable contribution, not as approved knowledge. A reviewer checks the cited evidence, corrects the proposal and an authorised Context Approver accepts the conclusion. The next Session begins with that decision and its evidence already available to the people permitted to use it.

What stays outside

Code editing, source control, test execution and issue management remain with the tools that already own them.

What cRUD Collab owns

Room purpose, permission-aware context, evidence relationships, working material, human approval and continuity.

What becomes durable

The approved decision, its evidence, its scope and the history of later challenges remain understandable beyond one meeting or chat.

This is the boundary we built for. Collaboration should connect the work without absorbing every tool, and organisational memory should become durable without making every contribution authoritative.

Summary

cRUD Collab is built to connect authorised people and approved AI tools to shared Product and Project Context, traceable evidence and human-approved knowledge. It provides browser-first Rooms and Sessions while preserving durable organisational memory, without trying to replace the editor, repository or tools teams already use.

Frequently asked questions

What problem led us to build cRUD Collab?
Teams could find the work in their existing tools, but the reasoning behind it often had to be reconstructed. We started with the problem of making that reasoning available when someone else needed to continue the work.
What should stay attached to a decision after the original discussion ends?
The evidence, the scope of the decision and the people responsible for it should stay connected. Those details help someone understand when the decision applies and when it needs to be revisited.
What is the durable value of cRUD Collab?
The durable value is permission-aware organisational memory that preserves approved decisions, evidence, guardrails and history for later people, sessions and approved tools.

Product status

cRUD Collab is in development.

The product has not been released. Explore the current design and product thinking, or contact the team about the direction.