01 — Typed records
The work has shape.
Notes, tasks, projects, people, organisations, and opportunities become records your agents can read, write, and reason over.
The agent-native operating layer for work
Your agents start from zero. Mine don't anymore.
Your agents need more than memory. Core Edge gives humans and agents one system for notes, tasks, projects, people, opportunities, blockers, links, and next actions.
Your data. Your markdown. Your call.
fig. 01 — the core, observed
notes · tasks · projects · blockers
state linked next visible
The problem
Your agents don't remember yesterday. Neither do you, honestly.
You paste the same project context into your agent every morning. The notes are in Obsidian. The tasks are in Linear. The client context is in your CRM. The real decision is buried in a chat thread. Client A's requirements are bleeding into Client B's project. You can't tell what's next, what's blocked, who owns it, or which source of truth to trust.
Your Obsidian vault is a graveyard of good intentions. Notion lasted three weeks. This time it'll be different. Right?
Core Edge is what I built when I accepted that agents don't just need notes. They need work state they can read, update, link, and trust.
Why I built this
Second brain? Try sixth. I had a problem.
Obsidian. Notion. Apple Notes. Raycast Notes. Markdown files scattered across projects. Nothing was findable. Everything was somewhere.
I kept rewriting the same context from memory because I couldn't trust any of the systems I'd built to hold it. What I actually needed wasn't better organisation — I needed somewhere to speak to my brain. A place where every idea, every scribble, every half-formed thought could be found fast and would never get orphaned.
So I built it. And then something unexpected happened: I stopped writing my own notes. I brain-dump now. The agents help curate.
Then I realised memory wasn't enough. Agents didn't just need notes. They needed tasks, projects, people, opportunities, blockers, status, links, and a way to know what mattered next.
This is the system I actually use every day.
The primitives
The primitives of agent-native operations.
Existing tools store fragments of work. Core Edge gives agents the primitives they need to
operate: records, lifecycle, links, and a trustworthy next step.
02 — Operational links
The graph is not decorative.
Tasks belong to projects. Blockers point to tasks. People own work. Status changes matter. Links affect what the system recommends.
03 — What next
Stop reconstructing the queue from memory.
what-next surfaces ready work. space review finds lost context, duplicates,
and stale threads before you waste the first hour remembering.
What it actually looks like
The moment it clicks.
You've just returned to a messy work queue after a week away. Notes, tasks, blockers, and
client threads all exist somewhere — but you need to know what to do next.
you
What should I work on?
→ core-edge what-next
agent
The next actionable task is the landing page rewrite. It's active, high-context, and
unblocked. I suggest landing that before opening a new thread.
you
Can I trust that queue?
→ core-edge space review
agent
Mostly. One active task is orphaned, and there are no stale blocker links. The queue is
safe enough to act on.
you
Show me the project context.
→ core-edge project context <id>
agent
Here are the child tasks, blockers, and status counts. Suggested next action: finish the
copy pass, then validate the page before touching implementation work.
No spreadsheet archaeology. No prompt archaeology. Just navigate.
Honest answers
I tried everything too.
On Obsidian
Beautiful. Mine is 400 notes deep and I haven't opened it in months.
Obsidian is a vault of markdown files on your disk, designed for one human. Agents can read and edit those files — they just weren't the audience when it was built. No per-agent identity, no retrieval built for machines, no transactional backend when multiple agents write at once.
Core Edge can import and export markdown. The hosted app is the primary alpha experience; local vault sync remains an operator workflow with known filesystem constraints.
On Notion
Great for teams. My agents treated it like a foreign country.
Notion has great APIs. An MCP adapter works. What it isn't is a protocol — agents talk to it through a layer built for people. It works until you need to know which agent changed what, or why something disappeared.
Core Edge was built agent-first from day one. Every read, every write — attributed.
On just using a local vault
I tried pointing my agents at my local files. Surely a quick grep would do it. Right?
One agent, one machine — fine. Then you add a second agent. Then your vault hits a thousand files. Then grep returns 60 results and none of them are the right one. Then you spend 20 minutes narrowing it down manually. Then you give up and rewrite the context from scratch.
Sound familiar?
Core Edge is what I bolted onto my workflow when I accepted none of this was working. Linear is excellent execution infrastructure; Core Edge gives agents durable context, links, people, projects, blockers, and rationale underneath agent-assisted work.
Trust & control
Your knowledge. Clear access. Portable records.
Source of truth
Markdown. Always.
Your notes are plain markdown files. Everything exports. Nothing is locked in a proprietary format. Whatever happens, your knowledge is yours and it goes where you go.
Activity
Nothing is hidden.
Every agent action is logged. Every read, every write, every change — attributed to the agent that made it. You always know what happened and you can always undo it.
Training data
Your data is yours.
We do not train models on your content. Hosted operator access is limited to running and supporting the service, and platform telemetry does not expose your record content.
Deployment
Start with the hosted alpha.
Who it's for
You'll know if this is for you.
01
Agentic engineers.
You run agents — Claude Code, Codex, Pi, custom scripts — and every one starts from zero. You've tried Obsidian, Notion, Roam, Logseq. You have 400 notes and no system. You paste context into prompts like it's your morning commute.
02
Founder-led businesses.
You are running clients, projects, opportunities, and delivery with agents in the loop. You need one place where work state survives the chat window.
03
Product teams building agents.
You need agent workflows inside a product, not a pile of prompts around it. Namespaces let different products, teams, or clients operate from separate work graphs.
Probably not for you if you haven't felt the pain of explaining
your own project to your agent for the 50th time. Come back when you're as frustrated as I was.
How it works
If you're curious.
A Rust service sits between your agents and a Convex backend. Markdown files are the permanent record. Convex is the live operational layer. The edge service handles auth, retrieval, and the graph.
The stack is opinionated so you don't have to be. You get fast retrieval, full attribution, and markdown files you can take anywhere — without becoming an infrastructure engineer to get there.
Start free
Create your account. Join the alpha queue.
Create and secure your account now. Core is in closed alpha, so operational access opens after your application is approved.
Durable free account Recoverable alpha profile No payment details
Create your free account
Already have an account? Sign in
FAQ
Questions worth asking.
How is this different from Obsidian or Notion?
Obsidian is a local markdown workspace and Notion is a SaaS built for human teams. Core Edge was designed around agents and humans sharing scoped access, retrieval, typed work, links, and attributed activity. Markdown import and export preserve portability, but the hosted app is the primary alpha experience.
Can I import my existing Obsidian vault?
Core Edge can import markdown with frontmatter, export records as markdown, and run its existing local sync daemon against a normal local directory. Obsidian and iCloud workflows still have sharp edges, so they are not the recommended alpha onboarding path.
Can I self-host?
Not as a self-serve alpha option yet. We are designing a supported deployment path for operators who need their own service and backend. Select self-hosting interest in the application so we can learn your requirements.
Does this replace Linear or my CRM?
Not today for most teams. In alpha, Core Edge works best beside your existing tools. But the primitives are intentionally broader: tasks, projects, people, organisations, opportunities, blockers, status, links, and context. Over time, some users may use it as their primary operating layer.
What does the hosted version cost?
Alpha is free. Hosted pricing will be announced before general availability. We are not optimizing for pricing signal right now.
Is my data used for training?
No. We do not train models on your content. Hosted operator access is limited to running and supporting the service, and platform telemetry does not expose customer record content.
When does the alpha open?
Rolling. We accept testers in small batches as we harden onboarding.