An agent that does the work across your documents, and waits for your sign-off.
Mindecho puts an agent inside your workspace that writes, edits, and runs multi-step work across your files. Every change is a diff you accept or reject. Nothing ships without you.
Free for personal use. No card. Your notes stay plain markdown you can export anytime.
The models got smart.
Your tools didn’t.
Today you paste text into a chat, paste the answer back, and hope it didn’t break the rest of the document. Or you hand an agent your files and hope harder. One is tedious, the other reckless. Neither does the work.
A different shape for agents at work
In your workspace, not a tab away
The agent reads and writes the same documents you do, side by side. No copy-paste, no handoff to another app.
Across many documents at once
Change a policy and the agent updates the references, the changelog, and the procedures along with it, in one coherent set of edits.
You approve every change
Every edit is a diff. Accept it, reject it, or route it to the right teammate. You see every change before it lands.
Real work, not chat
You saw the diff. Behind it is an agent that works across everything you keep here, and shows you each step before it counts.
Intake
Turn an email into a tracked case
A member writes in. The agent reads it, flags the legal area and urgency, pulls out the key facts, and matches the sender to your records. Then it hands you a case to open in one click, with the email thread attached so nothing gets retyped.
Research
Draft from your own library
Save the books, PDFs, and pages that matter. Ask for a board briefing or a blog post on the infant gut microbiome, and the agent reads across everything you’ve collected and writes a first draft, sources cited.
Handoff
Hand a step to a person, and it waits
Some steps need a human. The agent routes the next one to the right teammate by their role, then stops. Nothing moves until someone signs off, and every approval stays on the record.
Orchestration
Run a whole pipeline, not one prompt
One instruction becomes many steps: gather the facts, write the copy, generate the image. You get a finished social post to approve or reject.
Reasoning
Consult across the documents you already have
Point it at your own files and it writes a reasoned recommendation, with every claim linked to its source.
Let agents do more without losing control
The approval layer is the point. An agent can touch many documents and many steps, but every change waits for a human. Assign approvals by role, see what changed before it’s real, and keep a record of who decided what.
Your data, your model, no lock-in
No black box. The parts power users ask about first, and how they work.
Any model, per chat
Pick Claude, DeepSeek, or an OpenRouter-routed model per conversation. Dial reasoning effort up or down, switch between acting and planning first, or run a chat on a local LM Studio model.
A council of agents
Assemble expert agents, each with its own instructions, context, and model, then have them weigh in on a question in turn before you decide.
Threads that persist
Every conversation is saved and resumable, with its case and library context intact.
Approvals by role
Route any change to the right teammate to sign off, addressed by their role, not their name.
Plain files, no black box
Import your markdown and PDFs, export anything, with a lossless converter that round-trips it all.
Web and desktop
Use it in the browser, install the Mac, Windows, or Linux app, or clip from anywhere with the extension.
Email and calendar, in one place
Connect Gmail or Outlook, read and reply with agent-drafted replies you approve, and schedule events linked to the documents they belong to.
Who it’s for
A second brain that writes back
Capture everything, think out loud, and let the agent turn notes into finished work. Free.
Start freeWork that has to be right
Triage requests into cases, route approvals by role, and share a library and calendar, with an agent that never ships without sign-off. From €12 per seat.
Book a demoBe one of the first to put an agent to work.
Mindecho is rolling out in waves while we build it in the open. Personal use is free. Teams and small orgs get hands-on onboarding as we bring them on. Leave your email and we’ll send an invite.
- Governed by design — every change waits for a human, with approvals logged by role.
- Your data stays yours — plain, exportable markdown, with self-hosting on the roadmap.
- Built in the open — follow the build and help shape what ships next.
Build in public
Build in publicMindecho is built the way it’s meant to be used: one founder running the whole company with an army of agents inside it. The company is the demo.
Follow the buildGive an agent your documents. Approve what it does.
Your data stays plain, exportable markdown.