In development. Early access opens to a small group first.

Start each day with the right work.

ShipBench reads your analytics, your search data, and the marketing work you never finished, then hands you the few actions that matter today, with the reason behind each one. Built for solo founders who do their own marketing.

One email when your spot opens. Nothing else.

The Today bench, as it stands. The product is in development, so this will keep changing.
Up to three actions per product, not a backlog Every action explains itself Nothing goes public without your say

The problem

When you are the whole marketing team, deciding is the job nobody scheduled.

Analytics in one tab, Search Console in another. A half-written LinkedIn post, a Reddit thread you meant to answer, a landing page untouched since launch. None of it tells you what matters today. So the morning goes to deciding, and the doing slips to tomorrow.

Dashboards report. They don't decide.

Every tool shows you numbers. Not one of them turns a page losing clicks into a task with a reason and a deadline.

The plan lives in your head.

Which means it resets every morning, and it gets outvoted by whatever is loudest in your inbox.

Drafts pile up. Nothing ships.

Ideas, half posts, and good intentions sit in a notes app while the product's growth waits on them.

How it works

One loop. Run every morning.

Signals in, a short plan out, your approval on anything public, and the result measured so the next plan is sharper.

  1. 01

    Connect your evidence.

    Google Analytics, Search Console, PostHog, your email tool. ShipBench reads them. It does not replace them.

  2. 02

    Set one goal per product.

    “Open early access to the first group of founders.” “Turn free users into the first paying ones.” Every action is judged against that line.

  3. 03

    Get today's bench.

    At the hour you choose, ShipBench picks up to three actions per product. Small enough to finish, specific enough to start.

  4. 04

    Do, defer, or dismiss.

    Each action carries its reason and its expected outcome. Publishing and sending wait for your approval unless your policy says otherwise.

  5. 05

    Results feed tomorrow.

    What shipped, what moved, what didn't. The next bench is built on that, not on a template.

Evidence, not vibes

Every action explains itself.

Open any action and you get two lines: the reason ShipBench chose it, and the outcome it expects. If the reason is weak, dismiss it. The plan gets better when you do.

  • A reason tied to a signal, a goal, or a piece of unfinished work
  • An expected outcome you can check next week
  • Defer to tomorrow, or dismiss, in one click

Today's bench

What needs your attention

1 of 3
Website Assistant for Basecamp

Refresh the pricing page before the next LinkedIn post goes out

The post points people at pricing. Make sure the page answers the question it raises.

▾ Why this, now?

Reason Search Console shows the page earning impressions for a query it never answers. The post you approved will send more people to it.

Expected Visitors who arrive from the post find the answer on the page, and trial starts from it are measurable next week.

Done Tomorrow Dismiss this action

Account

Human-in-the-loop policies

  • Publish a LinkedIn or X post Require approval
  • Send an email campaign Require approval
  • Change a live page Require approval
  • Draft replies and posts Autonomous
  • Research and prioritize Autonomous
  • Weekly review and report Autonomous
1 approval waiting: LinkedIn post, drafted 18:40 Review →

Human in the loop

It prepares the work. You decide what leaves the building.

Research, prioritizing, drafting, and internal updates can run on their own. Publishing a post, sending an email, or changing a live page stops at an approval gate. You set the policy per action type, and you can loosen it as trust builds.

  • Default: approval required for anything public
  • Per-type policies, autonomous or require approval
  • Every run logged, every approval recorded

Claude and Codex

Use it from your own AI client.

ShipBench is also a remote MCP server. Connect Claude or Codex with a scoped token and the agent reads your strategy, your plans, and today's bench, then acts through the same permissions and approval gates as the app.

  • Scoped tokens you can revoke
  • Workspace-level authorization on every call
  • Same rules whether the action comes from the app or the agent
shipbench.ai/mcp scoped token · 2 products

What should I work on today for Assistant for Basecamp, and why?

shipbench.today project=assistant-for-basecamp

shipbench.document strategy.md

Two actions are on your bench, both tied to the MRR goal in strategy.md:

  1. Publish the voice-transcription post. Approved yesterday, still unpublished. It's waiting for your go-ahead.
  2. Answer the Reddit thread asking for a Kanban view. A real question you can answer from experience.

Publishing needs your approval. Say the word and it goes out through ShipBench.

Workspace

Export .zip
  • assistant-for-basecamp/
  • strategy.md
  • goals.md
  • playbooks/
  • linkedin.md
  • seo-refresh.md
  • plans/
  • today.md
  • drafts/
  • voice-transcription-post.md
  • reports/
  • week-34.md

Your workspace, readable

Strategy, plans, and drafts as Markdown you can actually read.

Each product gets a workspace: goals, strategy, playbooks, daily plans, drafts, and reports, all as Markdown documents. Open them, edit them, export them. Nothing that matters is locked in a table you can't see.

  • One workspace per product, each with its own goal
  • The same documents your AI client reads over MCP
  • Export everything, any time

Where the actions land

The channels you already use. Not a new one.

ShipBench connects the tools you already pay for through their public APIs. When a tool has no usable API, it prepares the work and walks you through the step.

  • LinkedIn posts and comments
  • X posts and replies
  • Reddit answers in the right threads
  • Website new pages, SEO refreshes backed by evidence
  • Email setup and guided sends

Boundaries

Three things ShipBench is not.

Not an analytics dashboard.

It reads yours. It doesn't draw new charts. The output is a task, not a graph.

Not a project manager.

It produces a few actions a day and expects them done. It does not want your whole backlog.

Not a social scheduler.

It drafts, you approve, and it posts through the tools you already use. No new queue to babysit.

Why it exists

Built on a routine I already run.

I run Assistant for Basecamp, a paying product, and I do all of its marketing myself. The routine is the same every time: pull the numbers, decide what to ship, draft it, post it, check what moved. ShipBench is that routine turned into software. Assistant for Basecamp is the first product on the bench, and nothing opens to other founders until it works for mine.

AR

Asad Ur Rehman

Founder, Karobar Solutions

FAQ

Questions founders ask first.

What is ShipBench, in one paragraph?

A growth operating system for a founder who does their own marketing. You connect your analytics and search data, set a goal per product, and at the hour you pick ShipBench hands you a short list of actions with the reason behind each one. You do them, defer them, or dismiss them. Anything public waits for your approval. It runs as a web app and as a remote MCP server for Claude and Codex.

Is this another AI wrapper?

The model is the least interesting part. ShipBench's job is the loop around it: pulling real signals, holding them against your goal, producing a small plan, gating what goes public, and measuring what happened so the next plan is better. The AI provider is a setting. You can run the planner on simple rules, on Cloudflare Workers AI, or on your own OpenAI key.

What does it connect to?

At launch: Google Analytics, Google Search Console, PostHog, and an email provider. More sources will follow where the tool has a public API. Where a tool has no usable API, ShipBench prepares the work and walks you through the step instead of pretending to automate it.

Will it post or send things without me?

Not unless you tell it to. The default policy requires your approval for every public action: publishing a post, sending an email, changing a live page. Research, prioritizing, drafting, and internal updates can run on their own. You set the policy per action type and can loosen it as trust builds.

How does it work with Claude or Codex?

ShipBench exposes your workspace and today's bench over MCP. You create a scoped token, connect your AI client, and the agent can read your strategy, plans, and actions, then act through the same services and approval gates the web app uses. Revoke the token and access ends.

What does it cost?

Pricing isn't decided. Early-access founders will see it first and get a say in it. There is no card to enter today.

When can I use it?

ShipBench is in development. Early access opens to a small group of solo founders first, and the only promise on dates is that you'll get one email when your spot is ready.

Who is it for, and who is it not for?

For a solo SaaS founder responsible for their own marketing, with a product that is live or about to be. Not for agencies or marketing teams, at least not yet. The product is built around one person making the calls.

Get on the early-access list.

Early access opens to a small group of solo founders first. Leave your email and you'll hear from Asad when your spot is ready, and at no other time.

One email when your spot opens. Nothing else.