LaunchRepoMost articles about directory launches assume a budget: a featured slot here, a “skip the queue” there, a paid plan on the review sites. This one assumes the opposite. The rule for our own seomap.dev run was zero spend — not “cheap”, zero — and the LaunchRepo playbooks inherit that rule: the agent is told never to spend money, start card-backed trials or accept paid upgrades. What follows is what a launch looks like under that constraint: slower, with more waiting and more handoffs, and with a tracker that stays honest about it.
The no-spend rule, stated precisely
In AGENTS.md, the repository’s instructions to the agent, the rule is short: never spend money or accept unrelated offers; do not start card-backed trials; do not accept paid upgrades. It sits next to the other hard rules — no invented customer counts, no manufactured votes, no evading account limits — because it is the same kind of rule: a boundary the agent must not cross even when a form makes crossing it the path of least resistance.
Two consequences. Any directory whose only usable route costs money is recorded as deferred_paid — nothing bought, the route noted, the decision left to you. And any upsell that appears inside a free flow has to be declined explicitly, which is more work than it sounds.
Declining upsells is a step, not an afterthought
Free routes are rarely one button. In our run, the free path on several platforms went through an offer, a decline, a confirmation and a second offer before the success page. The public field guides document what we saw:
- TinyLaunch: the Standard launch offered optional services, which had to be set to none; after scheduling, a second upgrade dialog appeared and the free path continued only through a “continue with free launch” option. The TinyLaunch guide lists the sequence.
- SaaSHub: the submission plan had to be inspected before submitting; the observed free option is documented as an observation, not a permanent promise. The SaaSHub guide covers it.
- PitchWall: the available plan is shown before the URL import; the PitchWall guide notes reviewing it, together with login consent and newsletter requirements, before proceeding.
The zero-spend planning guide turns this into a rule: review every final screen. Check the actual total, check the final confirmation, and reopen saved fields to verify they persisted. A queue jump you did not mean to buy is still a purchase.
No queue jumps, no workarounds
Zero budget tempts people toward things that are free but wrong. The playbooks forbid all of them, by name:
- Buying a queue jump — out, obviously. But also: re-submitting to move up a queue. The runbook tells the agent to recheck the existing record rather than submit again.
- Creating a second account to get around a limit. Uneed’s observed free workflow allowed one pending launch per account and blocked a new submission while a product was waiting. The Uneed guide records that as blocked with a next action, not as a problem to route around.
- Deleting another product to free a slot. Same answer.
- Asserting eligibility you have not verified. Fazier’s observed basic route required community comments, a visible badge and a domain rating above zero in a named checker. A product that does not meet that stops before attesting compliance — blocked, not “close enough”. The Fazier guide is explicit.
None of this is moralising. Directories that catch workarounds remove listings, and a removed listing after weeks of waiting is the most expensive outcome a free run can have.
Human handoffs are where the time goes
A free-only run has more handoffs than a paid one, because the paid routes are often exactly the ones that skip a review, a badge or a verification. The agent stops and records prepared_needs_human with the exact next action for:
- CAPTCHAs and two-factor prompts. Never bypassed, never solved by the agent. You complete them in the browser you connected.
- Website badges. Several free tiers require the directory’s badge on your site. That is a change to your product’s website, so it needs your recorded authorisation — once per product — and official badge assets. Refusing is fine; it simply rules those directories out.
- Verification emails. Registrations finish by email. If the agent has mailbox access in the connected browser, it reads the code itself; if not, you forward it, and the tracker records a human handoff rather than automated access.
- Decisions about paid routes. deferred_paid rows wait for you. Sometimes the answer is “no, skip it”; sometimes a directory matters enough that you pay once, outside the run, and record it yourself.
The local dashboard groups all of these under “needs you”. On a zero-budget run, that list is the real to-do list; the agent’s part is mostly done when it appears.
Review times: weeks, not days
The single biggest adjustment is time. Free tiers queue. The onboarding session in the repository makes the agent say this plainly before the first submission: many free-tier directories queue submissions for review, and confirmation or publication can take several weeks after submitting, not days.
Concrete observations from the field guides: PitchWall’s completed submission remained Under Review with an estimate exceeding 30 days. TinyLaunch confirmed a weekly slot with approval still outstanding. SaaSHub published a page that stayed labelled Pending approval. None of these are complaints — they are what free review looks like — but a tracker with one “done” column would have counted all three as finished on day one.
This is why product.json carries an earliest_launch_date and why the runbook tells the agent not to invent a platform SLA: if a directory gives no review timing, you agree a reasonable follow-up date and check then. The submission states article goes through the exact states and why scheduled and pending stay apart.
What our own pilot shows
The public evidence for all of this is our own product, seomap.dev, and it is deliberately limited to what you can verify. The current run has submitted seomap.dev to 160+ directories; each one is either live or waiting for the platform to publish. The publicly verifiable listings are linked from the about page. We do not publish account URLs, emails or screenshots of mail, and we do not publish a “live” count that includes pending pages.
Six of those platforms are documented in public field guides, and they show the whole spread on their own: Product Hunt reached scheduled, SaaSHub stayed submitted_pending_review behind a pending label, TinyLaunch ended scheduled with approval outstanding, PitchWall went submitted_pending_review with a 30-day estimate, Uneed came back blocked on an account limit and Fazier blocked on an eligibility rule. That distribution, not a single number, is the honest picture of a zero-budget run.
Budgeting attempts when the budget is zero
“Zero budget” still has a cost: the toolkit, once, and your agent’s usage, billed by your provider. The zero-spend guide suggests dividing the toolkit cost by products times planned attempts. The example it uses — EUR 100 across four products with six attempts each is about EUR 4.17 per attempt, excluding agent costs — is arithmetic, not a forecast. If only three directories fit your product, the allocation doubles. None of it predicts acceptance, traffic or ranking.
What zero budget buys you is a clean answer to the question every directory asks eventually: “did you pay for this?” No. Every listing in the tracker got there through the free route, within the platform’s rules, with a human stepping in wherever the platform wanted one.
Running your own
If you want to run the same constraint: the indie hackers page describes the one-person setup, the zero-spend planning guide is the checklist the agent follows, and the Claude Code walkthrough shows a full session including the handoffs. The repository — 346 playbooks, product template, tracker, dashboard — is a one-time purchase on the pricing page, and the playbooks will never authorise a purchase on your behalf.
Sources: AGENTS.md, ONBOARDING.md and RUNBOOK.md in the LaunchRepo repository; the public TinyLaunch, SaaSHub, PitchWall, Uneed and Fazier field guides (observed 2026-09-13); the zero-spend planning guide. The 160+ figure is the pilot’s own submission count as of this article’s date; it counts submissions, not live listings.