Join the waitlist for the next waveThis wave of the closed beta is full. Drop your email and you're first in line when the next one opens.Enter your emailJoin the waitlist✓ You're on the list! We'll be in touch soon.Please complete the security checkYou're on the list.You're #{position} in line.We'll email you the moment the next wave opens. Keep an eye on your inbox.Beta's full, join the waitlistJoin the waitlistJoin our free, limited closed beta.{filled} of {cap} beta spots filled· {waiting} waiting
For developers & power users

Altmind is the agent harness for your mind

A raw model is pure intelligence with nowhere to put it: no memory, no order, no hands. A mind needs somewhere to dump, something to organize it, a way to find it by sight, and the follow-through to act. Altmind is that harness, built around how your mind works. Bring your own model and drive it in plain language or a full REST API.

Visualize everythingControl & replayReady out of the box
altmind · dashboard
3
models connected
12
active skills
48
runs today
Automations
inbox-triagelive · byom:gpt-4o-mini
daily-briefnext 07:30
weekly-reportqueued Fri 17:00
Why we built it this way

Built around how a mind actually works

Your mind is brilliant at having ideas and terrible at holding them. A raw model adds intelligence, but a mind needs four things around that intelligence. So we built them in.

01

A mind needs to dump

Relief comes the moment a thought leaves your head. Altmind takes it any way it arrives, a sentence, a voice note, a forwarded email, with no forms and no filing. You think in bursts, and it catches every one.

Voice note
Forwarded email
Quick thought
Captured to memory
02

A mind needs organization it didn't do

Dumping only helps if something sorts it. Altmind files, links and tags every item the moment it lands, so structure appears without you building it.

#deadlineTimezone fix
#alexReply to Alex
#invoicePay hosting
Linked and filed automatically
03

A mind needs to find it back by sight

You remember where, not what. So Altmind gives your life a place: one dashboard where tasks, people and notes sit in view. You find things by looking, not searching.

Your day, in one view
Tasks6
Notes12
People4
Projects1
04

A mind needs to let go

A loop only closes when something happens. Altmind doesn't just remind you, it acts, follows through and reports back, so the thought leaves your head for good.

Renew insuranceopen loop
Renewed, confirmation filedhandled
Bring your own model

Swap the model, keep the harness

Connect the model you already pay for, OpenAI, Anthropic, Gemini, or a local one. Under the harness they are interchangeable parts: Altmind keeps the memory, tools and orchestration while inference runs on your key, so your automations are never capped by included tokens.

  • Swap models per skill: cheap one for triage, frontier one for drafting.
  • Keys stored encrypted; revoke anytime.
  • Run skills and automations as hard as you want: your key, no usage limits.
OpenAI
gpt-4o · o-series
Anthropic
claude family
Gemini
1.5 · flash · pro
Local / self-host
ollama · vLLM
BYOM pricing

Bring your key, pay less

When you connect your own model, inference runs on your key, so Altmind costs less in BYOM mode. Pricing is set after our closed beta.

Core
Set after the beta
Get beta access
Plus
Set after the beta
Get beta access →
Pro
Set after the beta
Get beta access

Prefer fully managed AI, no key required? See standard pricing →

Skills & automations

Describe a workflow. The harness compiles and runs it

Plain language in, working automation out. Each skill is versioned, observable and triggerable: on a schedule, an event, or an API call. No scripts to maintain.

  • Triggers: schedule, webhook, inbound message, or API.
  • Steps run with shared memory and your connected tools.
  • Every run logged: inspect, replay, roll back.
Trigger · 07:30 daily
Cron schedule
when
Read inbox + calendar
Gmail · Google Calendar
do
Draft brief + flag replies
model: byom
think
Send to Telegram
delivered 07:31
out
API · MCP

Drive Altmind from anything that speaks HTTP or MCP

A full REST API plus a hosted MCP server expose your memory, skills and automations, so Claude, ChatGPT, Gemini, or your own agent can read and run them.

GET/v1/memory/search
POST/v1/skills/:id/run
GET/v1/automations
POST/v1/automations
GET/v1/runs/:id
~/.config/mcp.json
"altmind": {
"command": "npx",
"args": ["-y", "@altmind/mcp"],
"env": {
"ALTMIND_TOKEN": "sk_live_…"
}
}
 
# now any MCP agent can call:
→ altmind.run_skill("weekly-report")
The dashboard

The harness, made visible

See what's connected, what's scheduled, and what ran, without leaving a chat. The dashboard is the visual layer over the same API you build against.

  • Live run log with inputs, outputs and replay.
  • Per-skill model + budget controls.
  • Token usage and cost across providers.
daily-brief
ran 07:31 · 1.2k tokens
ok
weekly-report
scheduled Fri 17:00
queued
inbox-triage
model: byom:gpt-4o-mini
live
The harness is usually code

Everything you'd wire by hand, already wired

Memory, a loop, tool brokering and observability: the four parts of a harness. Build them yourself, or get them managed and reachable from one place.

 
Build it yourself
Altmind
Memory across tools
Build & host a vector DB
Built in
Define a workflow
Write & maintain scripts
Plain language
Trigger from an agent
Custom endpoints
REST + MCP
Observe & replay runs
Roll your own logging
Dashboard
Recipes

A few things power users wire up first

Each is one plain-language skill: trigger, tools and model are yours to tune.

daily-brief

Morning command center

Summarize inbox + calendar, flag what needs a reply, and push one message before you wake up.

07:30 → read → draft → Telegram
inbox-triage

Always-on triage

Every inbound email gets labeled, prioritized, and, if routine, auto-drafted in your voice.

on email → classify → draft
weekly-report

Reporting on autopilot

Pull numbers from Sheets, write the narrative, drop a ready-to-share doc in Drive.

Fri 17:00 → Sheets → Docs
capture

Catch every thought

Forward a voice note, photo or link from anywhere; it's transcribed, tagged and filed to memory.

on message → extract → store
on-call

Alert router

Webhook in, Altmind decides severity, summarizes context, and pings the right channel.

webhook → assess → notify
research

Background research

Hand it a question; it gathers, cross-checks and returns a sourced brief to your vault.

ask → gather → brief

Become unstoppable

Plus and Pro unlock BYOM, the full API, MCP, and truly unlimited automations: connect your own key and run skills without hitting a token ceiling.

FAQ

Questions builders ask

Which models can I bring?

OpenAI, Anthropic and Gemini out of the box, plus any OpenAI-compatible endpoint, including local models via Ollama or vLLM. You can assign a different model per skill.

Is there really a full API and MCP server?

Yes. Every capability in the dashboard is backed by the REST API, and the hosted MCP server exposes your memory, skills and automations to any MCP-speaking agent.

Where are my API keys stored?

Encrypted at rest and used only to call your provider. You can rotate or revoke a key at any time, and runs that used it keep working until the next invocation.

Do I need to write code to build a skill?

No. Describe the workflow in plain language and Altmind compiles it. Power users still get full control via the API and per-step configuration when they want it.

Can I self-host?

You can keep inference local by pointing Altmind at a self-hosted model. Org-wide and isolated deployments are available on Elite; reach out for details.

The harness, without the wiring

Bring your model. Build a skill in a sentence. Run it from anywhere.