Meterless
GAIA · V2.4.1 · MIT
GAIA · THE MIND · 01 / 05

A cognitive OS that
takes action.

Gaia is the cognitive operating system that lives on your desktop. It can research, see, remember, and operate across your computer and its own browser to complete real work.

See the memory
Brains12
Context WindowsBoundless
Providers7/2 Local
Memory TiersShared
CARD 03Gaia Gaia is the cognitive operating system that runs on your desktop. It can research, see, remember, and operate across your computer and its own browser to complete real work.
PLANNING
SECTION · 02 / 04

What happens when AI is given
it's own entire OS.

Research
Operate
Persist

Desktop control, persistent memory, agent orchestration, and live research together in one loop. Experience emergent AI behavior you can't replicate elsewhere. All with only natural language.

Gaia · capabilities

See for yourself.

01 AUTONOMOUS · MISSIONS

State a goal. Gaia plans, executes, and verifies it.

One instruction, end-to-end. Gaia decomposes the goal, picks the apps, runs the research, writes the artifact, and hands it back. Every step is inspectable.

TRY A GOAL
IDLE
M-4307 · gr_1747526400_b7c8d9 · ELAPSED 00:00
“Plan my weekend trip to Portland and book the basics.”
Turns
0/5
Time
0.0/7 min
Cost
$0.00/$1.60
    ARTIFACT
    portland-weekend.md · itinerary + bookings
    READY FOR REVIEW
    H-MEM SYNC DEMO SPEED?
    02 SCHEDULED · MISSIONS

    Set the cadence. Gaia runs while you sleep.

    Anything Gaia can do once, it can do on a schedule. Daily briefs, weekly recaps, monthly comp refreshes. Each is a real mission with history, drift detection, and a clean handoff to you.

    Week of Apr 13
    5 ACTIVE · 142 RUNS
    Mon
    Tue
    Wed
    Thu
    Fri
    06:00
    07:00
    08:00
    09:00
    10:00
    13:00
    15:00
    DAILY · ACTIVE

    Folder triage + changelogs

    Crawl what changed in your working folder today, classify by intent, draft a changelog entry in your voice. Nothing committed without your approval.

    SCHEDULE?
    0 13 * * *
    Daily at 13:00
    LAST 7 RUNS
    7 / 7 OKAVG 2.4 MIN
    MISSED RUN · DRIFT DETECTEDGaia was not available at the scheduled time. The mission will resume on its next cadence; the gap is recorded in the run log.
    NEXT RUN
    in 1h 47m
    Simulate Gaia offline
    shows how drift detection handles a missed run
    03 OBSERVER · VISION

    Watch what’s on your screen. Coach you in real time.

    Live screen perception runs through Observer, local vision models, and the shared WorldState. Gaia can see what you’re doing across any app, and quietly flag what you should fix before you ship it.

    SHOW ME
    STANDBY
    DRIFT 0.00 ?
    figma.com / launch-LP / hero-block GAIA OBSERVING
    Pages
    Cover
    Hero block
    Features
    Pricing
    Ship faster.
    From idea to launch in a week.
    Start free
    Inspect
    color: #f9e6d4
    color: #f9c590
    btn opacity 55%
    01 / CONTRAST Hero title against the gradient is 2.1:1. Push the title to ink — you’ll clear WCAG AA.
    02 / CTA “Start free” button is at 55% opacity. Primary CTAs should be fully opaque.
    03 / NAMING Hero frame is still named “Frame 17.” Want me to rename it to “hero-block”?
    AB VendorC MRRD RenewalE Notes
    1Datadog$4,8202026-08-14SOC 2
    2Honeycomb$2,1402026-06-022024-06-02
    3New Relic$3,5602026-09-10SOC 2
    4Splunk$48.2002026-07-22quoted
    5Sentry$6402026-04-30SOC 2
    01 / TYPO D2 reads “2024-06-02” — almost certainly meant 2026, same year as the rows around it.
    02 / FORMAT C4 uses a period as a decimal separator. Sheets is parsing it as text.
    03 / SUM Total MRR on this view is $11,160 once C4 is fixed. Want me to add the formula?
    To: marcus@board.vcDraft · 14:02
    Q1 update · revenue, hiring, the path to launch
    Hey Marcus, We had a great quarter. ARR up 38% QoQ and we shipped the v2 cognitive stack. Two senior hires landed: head of design from Linear, staff infra from Stripe. The one thing we’re still working through is the on-call rotation — we had 3 incidents this quarter and our MTTR isn’t where I want it to be. Let me know if you want to grab time next week to walk through the road map. Best, Anita
    01 / TONE Subject mixes celebration and concern. Lead with what they want to know first.
    02 / NUMBER H-MEM has Q1 ARR growth at 42%, not 38%. Want me to swap it?
    03 / FOLLOW-UP You agreed at last board meeting to give MTTR a target. Add one before sending?
    14async function fetchPlan(goal) {
    15 const res = await fetch('/api/plan', {
    16 method: 'POST',
    17 body: goal // raw object, not stringified
    18 });
    19 return res.json();
    20}
    21
    22// usage
    23fetchPlan({ prompt: 'enrich leads' });
    01 / BUG Line 17: body is the raw goal object. Server expects JSON — wrap with JSON.stringify.
    02 / HEADERS No Content-Type header on the fetch. Production handler will return 415.
    03 / TEST No test covers this path. Want me to scaffold one in /tests?
    04 BROWSER · GROUNDING

    Its own browser. Real research, cited and structured.

    Gaia spins up a native browser session, navigates real sources, captures live pages, and pulls verifiable facts into an evidence trail.

    gaia.local / browser / swarm-ai-labs
    GAIA BROWSING
    INTENT COMPARISON “Three research labs shipping swarm AI, with model scale and pricing.”
    CRUNCHBASE · CATEGORY VIEW
    Research labs building swarm AI
    Meterless, Anthropic, and Perplexity are the three labs publishing peer-reviewed work on multi-agent swarm orchestration this year.
    METERLESS · RESEARCH
    Meterless Swarms technical report · published Mar 2026
    Meterless’s Swarms coordinates up to 125 agents on a shared blackboard, with a published planner-merger-verifier loop and open evaluation harness.
    KIMI 2.6 SWARMS · PRICING
    Research API tiers
    The Research tier includes 128-agent swarms, weight access for academic use, and dedicated support. Enterprise pricing is quoted by request.
    PERPLEXITY · COMPUTER
    Benchmark scores across the shortlist
    On the multi-agent track, Meterless leads on resolved-task rate while Perplexity wins on cost-per-task. Collective AI ranks highest on coordination efficiency.
    QUERY
    “Three research labs shipping swarm AI, with model scale and pricing.”
    0 CITATIONS EVERY CLAIM LINKED
    05 DESKTOP · CONTROL

    Native input across real apps. Not an API, the actual interface.

    Through a window session-based desktop model, Gaia moves the cursor, types, scrolls, and verifies state inside one app at a time using, no passwords handed over and no root access required.

    Gaia — Interview Prep Workspace
    CONNECTED · OAUTH, NO PASSWORDS
    LinkedIn OpenAI Google Drive
    8 EXECUTIVES
    A. Okafor · CFO
    PROFILES + USER STORIES
    Profile draft
    User stories
    Save to Drive
    ACTION LOG
    da_idle
    SESSION · IDLE
    TASK Research 8 execs before the interview — build a profile and user stories for each, save to Drive.
    0 ACTIONS RATE LIMIT 60/S
    SAFETY Gaia drives apps through your existing signed-in sessions — no passwords handed over, no root access. Control is bound to one window, every action audited and rate-limited to 60/s, and dangerous combinations (Alt+F4, Ctrl+Alt+Del) are blocked at the runtime.