Skip to content

Fewer tokens per verified outcome.

Track. Verify. Remember. Spend less.

Mark 1 runs in the background, learns from your agent's work, helps you verify outcomes, and turns experience into lower costs over time.

Available for
Works with
  • Claude Code
  • Codex
  • Gemini CLI
  • opencode
  • Copilot CLI
  • Cursor

How it works

How It Works

From agent activity to long-term improvement. A closed loop for more capable, efficient and reliable agents.

Layer 1

Session Traces

Capture the full agent session — attempts, tool calls, exploration and verification — and measure Tokens-per-Verified-Outcome.

AGENT |TOOLS |ENV |TRACES
Layer 2

Calibrated Adjudication

Combine weak signals into calibrated outcomes, even when the repository has no test oracle.

satisfied violated refused error
Layer 3

Procedural Memory

Turn verified experience into reusable workflows, pitfalls, invariants and routing hints — without carrying raw code or trajectories.

WORKFLOWS |INVARIANTS |ROUTING
Layer 4

Safe Retrieval & Learning

Retrieve useful experience, choose repair or resample intelligently, and admit shared artifacts only when their measured utility clears the safety gate.

RETRIEVE |REPAIR |LEARN |SAFE

Backed by

  • The Residency
  • Google for Startups
  • NVIDIA Inception
  • Aiven for Startups

More than memory

Deeper Visibility. Smarter Agents.

Everything you need to observe, verify, and improve your agents.

Trace Every Step

Observe

Capture every tool call, file access, and decision your agents make — with complete context.

Build a Procedural Memory

Remember

Turn successful runs into reusable, composable knowledge that your agents can recall and follow.

Verify Outcomes

Trust

Automatically check if the agent's result is correct, complete, and grounded in real evidence.

Find & Reuse Past Runs

Reuse

Search across all your agent runs using natural language. Instantly reuse what worked before.

  • Enterprise refund workflowUsed 12 times · 92% success 2 weeks ago
  • Stripe integration setupUsed 8 times · 100% success 1 month ago
  • Escalation to humanUsed 23 times · 87% success 2 months ago

Integrate With Your Stack

Connect

Works with the tools your agents already use. Zero changes to your existing workflows.

  • Claude Code
  • OpenAI
  • Cursor
  • GitHub Copilot
  • LangChain
  • Custom Tools

Reduce Costs Over Time

Optimize

Better memory means fewer repeated mistakes, shorter runs, and lower token usage.

Private · Verifiable · Built for a higher purpose

— Same agents. More intelligence.

Arrests

Stop at the knee, not at the bill.

Every retry is a bet. Mark 1 tracks the odds that the next attempt lands against what it will cost, and stops the loop when the bet is no longer worth it. You see the numbers it used.

The knee: chance the next attempt lands, against its cost The chance that attempt n lands falls with every attempt, from 0.71 on the first to 0.03 on the eighth. The cost of an attempt, as a share of task value, stays flat at 0.14. The curve crosses it at attempt 5, where Mark 1 stops. 0.80.60.40.2 12345678 attempt n stop c(n) / V p̂_LCB(n)
chance the next attempt lands cost of an attempt arrest
attempt
5
p̂_LCB
0.11
c(n) / V
0.14
Illustrative curve. Your numbers come from your own sessions.

Watch it work

The rule

p̂_LCB(n) · V(task) ≤ c(n)

When the lower bound on the chance that attempt n lands, times what the task is worth, is no more than what attempt n will cost, the loop stops. The verdict and the evidence go to you instead of into another retry.

Local caps, whatever the policy says

  • At most 3 blocks in a row
  • Hourly ceiling
  • Pause wins

See how much you can save

Pick the plans your team already pays for. Mark 1 sits beside them, arresting the retries that never needed to happen and routing the rest to the cheapest model that still verifies.

Your year, before and after

Wasted on retries

$9,600

You keep

$2,880

Stack covered 80%

recovered by arrests, memory and routing

  • Cursor $1,080
  • Claude Code $1,800

Mark 1 is $0 during the preview — arrests, memory and receipts included.

Get Mark 1 for macOS
  1. 01

    From decisionto evidence.

    Inspect the evidence behind Mark 1’s decisions.

    Inspect.Understand.Trust.

  2. 02

    Anchors

    Trustedevidencesources.

    Anchors are approved checks or tools associated with a project. They represent the evidence sources Mark 1 can use when evaluating work.

    More evidence.Better decisions.

  3. 03

    Gates

    Control whenand how.

    Gates determine when Mark 1 may stop, block, or intervene in a workflow. This includes stop-gate settings and related consent configuration.

    Safety built in.

  4. 04

    Arrests

    Stop whatdoesn’t help.

    Arrests show requests or retry loops that Mark 1 stopped because continuing was inefficient, unsafe, or below the configured threshold.

    Less waste.More signal.

  5. 05

    Verdicts

    Clearoutcomes.

    Verdicts summarize the result of evaluating a task or outcome. Review how verdicts were distributed and which evidence contributed to them.

    Every outcome.Backed by evidence.

  6. 06

    Tamper

    Protectwhat’s true.

    Tamper shows integrity problems or unexpected changes detected in the verification data. Identify when a result may no longer be trustworthy.

    Trust throughtransparency.

  7. 07

    A moretrustworthytomorrow.

    Verify connects Mark 1’s conclusions to the evidence that supports them. Inspect, don’t just accept.

Illustrative — mock data.

Memory

Knowledge
that stays.

Manage reusable knowledge from previous sessions.

Memory stores useful artifacts extracted from earlier AI sessions and makes them available when they can help with future work.

Explore Memory
  1. 01

    All your knowledge

    Admitted memory artifacts.

    See the knowledge captured from previous sessions and curated for future use.

  2. 02

    Understand the full picture

    Beyond what’s active.

    Browse extracted and retired artifacts, filter by type, and keep your memory clean and relevant.

  3. 03

    See memory in action

    Memory injections in real time.

    See when and where memory is injected into sessions, and what impact it made.

  4. 04

    Trace the impact

    Sessions and outcomes.

    See which sessions used a memory artifact, what the outcome was, and how much time (and tokens) it saved.

  5. 05

    Take control

    Manage your memory.

    Pin important artifacts, share with your team, retire outdated knowledge, and inspect exactly where each artifact was used.

  6. 06

    A faster, more consistent future

    Build on what you know.

    Memory helps AI agents build on previous work instead of repeatedly exploring the same codebase, making future sessions faster and more consistent.

Previous sessions. A smarter tomorrow.

Illustrative — mock data.

  1. 01 Route

    Understandhow AI requestsare routed.

    Route shows how Mark 1 chooses or evaluates model paths for individual requests. It explains model ladders, escalation decisions, routing policies, and the savings associated with using a lower-cost or more appropriate model.

    Explore Route

    One request.Manypossibilities.

  2. 02 Model ladders

    Start cheap.Scale onlywhen needed.

    A ladder is an ordered sequence of models. Mark 1 can begin with a lower-cost model and escalate when the task requires more capability.

    A ladderfor everytask.

  3. 03 The knee

    Know whento move up.

    The knee is the point at which Mark 1 determines that a task may need to move to a stronger model or a different route.

    Betterjudgements.Lower costs.

  4. 04 Route decisions

    See everydecision.

    Each route decision shows the task, the model used, whether it stayed or escalated, routing fidelity, probability, and estimated savings.

    Transparencyin everyrequest.

  5. 05 Routing insights

    Less spend.More progress.

    Route helps you reduce costs without compromising quality. See escalation rate, routing savings, and model usage distribution in real time.

    Realsavings.Realimpact.

  6. 06 Configuration

    Your rules.Your control.

    Configure exploration settings, view gateway and proxy status, check the applied routing bundle, and see the price table version.

    Flexibleenough foryour stack.

Mark 1

The right model for every request.

Lower costs. Higher quality. More control.

Illustrative — mock data.

Proof

Verify Omium. Don't trust it.

Every number on your dashboard traces to a receipt signed by your device and by our server. The verifier is a standalone binary that runs offline. Anyone can check a receipt. Nobody can edit one.

One receipt

receipt_185.json
" seq ": 185
" checkpoint ": 4
" issued_at ": 2026-09-18T14:02:19Z
" prev_receipt_hash ": b1f4a0c97d21…
" merkle_root ": 7d21c4e9a0b5…
" device_sig ": ed25519:…
" server_sig ": ed25519:…
" witness ": pending
offline · nothing leaves this page

Synthetic values, real field names. Hashes shortened for display.

The Mark 1 Receipts screen: four sealed checkpoints, each with device and server signatures and a witness pending chip, beside a Verify offline panel.
Receipts screen. Device and server signatures, witness pending hourly.
  • coverage band

    How much of the traffic Mark 1 actually saw. Every number carries it.

  • cost basis

    Dollars on API keys, quota on subscriptions. Never mixed in one report.

  • indicative or guarantee-grade

    Below the coverage floor a number is indicative and says so. Guarantee-grade is refused, not softened.

Agents

Works with the agent you already run.

Claude Code, Codex, Gemini CLI, opencode, Copilot CLI and Cursor. Mark 1 attaches through each agent's own hooks and config, never the binary. Pinned versions are listed, and anything it writes can be reverted byte for byte.

  • Claude Code

    Anthropic

    attach
    hooks in settings.json
    detach
    reverts byte for byte
  • Codex

    OpenAI

    attach
    hooks.json
    detach
    reverts byte for byte
  • Gemini CLI

    Google

    attach
    hooks in settings.json
    detach
    reverts byte for byte
  • opencode

    SST

    attach
    plugin
    detach
    reverts byte for byte
  • Copilot CLI

    GitHub

    attach
    env pointer
    detach
    reverts byte for byte
  • Cursor

    Anysphere

    attach
    hooks.json
    detach
    reverts byte for byte

Pinned versions are listed in the app.

Download · macOS · Windows · Linux

Install in under a minute.

Pick your platform. The first run asks for consent before it reads anything.

First run

  1. 01

    Sign in with Omium.

    One account for your team. Keys and sessions stay on your device.

  2. 02

    Pick the agents and projects it may read.

    Attaches through each agent’s own hooks — Claude Code, Codex, Gemini CLI and more.

  3. 03

    Consent to the checks it runs.

    Static checks on by default; test suites only where you turn them on.

Unsigned build

Signing is in progress. Until it lands, each OS asks for one confirmation on first open.

  • macOS 1 confirmation

    1. Open it once
    2. Privacy & Security
    3. Open Anyway

    Needed the first time you launch the app; every launch after is normal.

  • Windows 1 confirmation

    1. More info
    2. Run anyway

    Appears on the SmartScreen prompt; click through once and it remembers.

  • Linux one terminal line

    1. chmod +x Mark1.AppImage
    2. ./Mark1.AppImage

    Run both from the folder you downloaded to, then pin it like any app.

Pricing · Preview

Two ways your agent pays. Mark 1 knows which.

Subscription agents are priced in quota, so Mark 1 reports tokens and arrests, not dollars. API-key agents get routing too, and dollars appear only when the cost basis is real.

Regime 01

Subscription

price preview

Your agent draws on a plan quota. The plan picks the model, so there is no price per token to report.

  • Capture Session traces from the hooks you enable.
  • Verify Your tests, build, types and lint, recorded as verdicts.
  • Arrests Stops the retry loop when the next try is not worth it.
  • Memory Pitfall notes handed back at the same place next time.
  • Receipts Signed checkpoints you can verify offline.
  • Routing Not in this regime — the plan picks the model.

Reports tokens and arrests, not dollars.

Regime 02

API key

price preview

Your agent bills per token through your own key, so Mark 1 can see what every attempt costs.

  • Capture Session traces from the hooks you enable.
  • Verify Your tests, build, types and lint, recorded as verdicts.
  • Arrests Stops the retry loop when the next try is not worth it.
  • Memory Pitfall notes handed back at the same place next time.
  • Receipts Signed checkpoints you can verify offline.
  • Routing Per task type — the cheapest model whose measured pass rate holds.

Reports tokens, arrests and dollars. Dollars appear only when the cost basis is real.

Pricing arrives with the first guarantee-grade reports.

Real tasks. Measurable gains.

Benchmarks That Matter.

Mark 1 gives agents long-term memory and procedural recall, leading to higher success rates, fewer tool calls, and more consistent performance.

Long-Context QA

OpenBookQA

Answer questions that require information from earlier in the session.

+28% higher accuracy
Without Memory 42%
With Mark 1 70%

Fewer repeated lookups, better reasoning.

Tool Use Efficiency

ToolBench

Complete real-world tasks with fewer tool calls.

-40% tool calls per task
Without Memory 15.2
With Mark 1 9.1

Remembers past actions, avoids repetition.

Multi-Step Success

AgentBench

Higher success rate on multi-step agent tasks.

+32% success rate
Without Memory 46%
With Mark 1 78%

Builds and reuses procedural workflows.

Task Completion Time

Internal

Finishes complex tasks faster with learned procedures.

-35% time to complete
Without Memory 120s
With Mark 1 78s

Recalls proven paths, skips dead ends.

Benchmarks • Real-world tasks • Open evaluation

More memory. Higher intelligence.

FAQ

Frequently asked questions

Real questions. Straight answers.

Pricing

What does Mark 1 cost?

Nothing during the preview. At launch, Free stays free, Pro is $20 a month ($16 billed yearly) and Team is $35 a seat. Install during the preview and Pro is $12 a month for your first year. On an API key, if a month of receipts shows Mark 1 saved you less than the plan price, that month is credited. Nothing you install now changes when pricing lands.

Does routing work on my subscription plan?

No. On a subscription your plan picks the model, so there is nothing for Mark 1 to route. It still watches the session, gates retries and remembers what failed, and it still signs receipts. Routing needs an API key, because that is the only regime where Mark 1 can choose the model per task.

Why are some numbers marked indicative?

Until check coverage is above the floor and a noise certificate exists for your setup, Mark 1 cannot stand behind a number, so it labels it indicative. The app says so on every screen where such a number appears. Guarantee-grade numbers arrive once both conditions hold.

Agents

Which agents does Mark 1 support today?

Claude Code, Codex, Gemini CLI, opencode, Copilot CLI and Cursor. Mark 1 attaches through each agent’s own hooks or config file and never touches the agent binary. The pinned versions it has been tested against are listed inside the app.

What is an arrest, and can I turn it off?

An arrest stops the next attempt when another try is likely to cost more than it is worth. Mark 1 arrests at most three times in a row and never more than an hourly ceiling, and if you pause it, the pause wins. The off switch is in Settings.

What counts as a verified outcome?

An outcome your own tests, build, type checker or lint passed. Mark 1 runs the checks you consented to and records the result as a verdict: satisfied, violated, refused or error. No second model grades the first.

Privacy

What does Mark 1 read, and what does it send anywhere?

It reads only the projects you enable: the agent’s hooks, the tail of its transcript, and the anchors you consent to run. What leaves the machine is typed evidence, verdicts and receipts, plus sanitized diff hunks only if you opt in. Full diffs and source code never leave, and what does leave is pinned to your region.

What is a receipt, and how do I verify one offline?

A receipt is a checkpoint over your session events, signed by your device and by our server. The verifier is a standalone binary that runs offline, with no account and no network. Export a receipt from the Receipts screen and anyone can check it without trusting Omium.

Install

Why is the build unsigned, and how do I open it?

Code signing is in progress. Until it lands: on macOS, open it once (macOS blocks it), then System Settings → Privacy & Security → Open Anyway (on older macOS, right-click the app and choose Open). On Windows, choose More info, then Run anyway. On Linux, chmod +x the AppImage.

How do I remove it, and does it leave anything behind?

Quit stops the app; Stop Mark 1 completely, under Settings, stops the daemon too. Removing a project or an agent reverts every line Mark 1 wrote to that agent’s hook or config file byte for byte, and the app shows the diff before and after. Your code and transcripts were never copied, so there is nothing to delete elsewhere.

( let's talk )

Turn Agent Memory
Into Real Impact.

Talk to our team and see how Mark 1 can help your agents work smarter, remember more, and get more done.

Book a Call

MORE MEMORY. HIGHER INTELLIGENCE.

BUILT FOR WHAT'S NEXT.