Your projects
should remember.
Every engineer and AI agent learns something about your software. Nebulai makes sure the project remembers it — as queryable project intelligence for whoever works on it next.
As the project grows and matures, businesses spend less on tokens. Agents retrieve that understanding instead of reconstructing it every session.
Nobody has the whole picture.
GitHub knows what changed. Linear knows what was planned. Conversations hold why decisions happened. Developers know what they discovered. Each system only contains part of the project.
Nebulai turns those fragments into a continuously evolving model of the project — how it works, what’s changing, why decisions were made, what’s blocked, and what the team has learned.
Explore Project IntelligenceCONTEXT
How project intelligence compounds.
Nebulai isn’t another coding agent. The project is the unit of intelligence — not a generic company knowledge base.
Observe
Repositories, tickets, conversations, and history generate signals as the team works.
Understand
Nebulai connects those signals into Project Intelligence: architecture, decisions, dependencies, and activity.
Retrieve
Developers and agents pull the relevant Project Context instead of reconstructing it.
Discover
Work uncovers hidden behaviour, constraints, debt, and conventions that were never written down.
Remember
Important Discoveries become Project Memory, available to the next person, agent, or session.
Know the project before you touch the code.
When someone joins a project, they should already understand what it does, how it is built, what is active, what is blocked, and what the team has learned. Nebulai opens with that briefing — not an empty dashboard.
Good morning, Alex.
Orion is migrating authentication from Auth0 to Entra ID while the customer portal team prepares a new organisation onboarding flow. Three authentication PRs merged this week. Portal deployment is blocked pending production Key Vault permissions.
Every coding session can make the project smarter.
A Discovery is something learned while doing the work: hidden behaviour, technical debt, a security concern, an architectural constraint, or an unexpected dependency.
Nebulai captures it with evidence, associates it with the work where it emerged, and keeps it in Project Memory.
- Compare against existing Project Memory
- Preserve evidence without interrupting the task
- Keep confirmed Discoveries attached to the work that found them
Webhook retries have no maximum attempt limit
The retry worker continues rescheduling failed Stripe webhook deliveries indefinitely. This may grow the queue and repeatedly process permanently invalid events.
- Discovered while
- ENG-402
- Service
- billing-service
- Evidence
- retryWorker.ts:84
- Related PR
- #824
Stop paying your agents to relearn the same project.
Your project is bigger than a context window. Nebulai makes the relevant parts addressable — so Claude Code, Codex, Cursor, and future agents retrieve Project Context through MCP instead of reconstructing it every session.
78 async function processRetry(job: RetryJob) {
79 try {
80 await deliverWebhook(job.event);
81 } catch (error) {
82 await queue.reschedule(job, {
83 delay: exponentialBackoff(job.attempt)
84 });
85 }
86 }Understand the story, not just the documents.
Tickets, decisions, PRs, and events outside the repository still determine whether work can move. Nebulai keeps that story connected.
Investigate dropped Stripe webhooks
Exponential backoff, preserve delivery order
Billing worker gains retry queue
Improve webhook reliability
Retry worker has no maximum attempt limit
Production Key Vault access
Organisation onboarding cannot deploy until the Infrastructure team grants the production service principal permission.
- Related work
- ENG-519
- Owner
- Infrastructure team
- Requested
- Tuesday, via email
The project remembers, even when people move on.
Engineers leave. Slack threads get buried. Coding sessions end. Agents lose their context. Somebody forgets why a decision was made.
Nebulai turns that ephemeral understanding into durable Project Memory — attached to the project itself.
Self-deploy it.
Your codebase stays yours.
Run Nebulai in infrastructure you control. That is the security model: source, embeddings, and Project Intelligence stay behind your boundary — not in a vendor cloud.
Deploy it in your environment. Ingestion and retrieval do not require a Nebulai-hosted service.
Scope repository and integration access to what each project needs.
Access follows the project: who is on it, and what they can retrieve.
Know what is connected, who can access it, and how to disconnect it.
Keep your tools. Connect their context.
GitHub still owns the code. Linear and Jira still own planned work. Conversations can stay where they are. Coding agents still do the coding. Nebulai connects the knowledge those systems produce into durable project intelligence.
Your codebase has a history.
Give it a memory.
Bring a project into Nebulai. Connect its code, work, and engineering history, and see what changes when every developer and agent can start with the project’s existing understanding.