Give it a goal

curl -X POST "https://api.sb/leaseline.app/goals/pursue?mode=try" \
  -H "Authorization: Bearer $SB_TOKEN" \
  -H "Content-Type: application/json" \
  -d '{"objective":"Book 5 calls with property managers this month from cold email. Send only from the warmed domain.","by":"2026-11-26","budget":{"amount":40,"currency":"USD"}}'

npx startup-builder do Goal.pursue --try --data '{"objective":"Book 5 calls with property managers this month from cold email. Send only from the warmed domain.","by":"2026-11-26","budget":{"amount":40,"currency":"USD"}}'

{
  "name": "do",
  "arguments": {
    "mode": "try",
    "code": "return await $.Goal.pursue({ objective: 'Book 5 calls with property managers this month from cold email. Send only from the warmed domain.', by: '2026-11-26', budget: { amount: 40, currency: 'USD' } })"
  }
}

import { $ } from 'startup-builder'

const result = await $.try.Goal.pursue({ objective: 'Book 5 calls with property managers this month from cold email. Send only from the warmed domain.', by: '2026-11-26', budget: { amount: 40, currency: 'USD' } })

The reference for every call on this page is Goal mode.

Two goals, spent well under budgetCopy Anchor Link

A founder with no customers has two questions first: who do I write to, and what do I write about. These two goals answer them.

The first asked Build a verified list for 50 verified heads of operations, with a $4 budget. The second asked the Keyword opportunity map for 20 keywords a new site can win, with $1.

Each stopped the moment it had what was asked. A budget is a ceiling, never a target: the agent spends nothing to use it up. Each cent below has a row on its receipt.

50 verified heads of operations (or COO / VP Operations / Director of Property Operations) at US residential property management companies managing roughly 100–1,000 units.

Achieved

50 of 50 verified

Spent

$0.29 of $4.00

Calls

136

Agent time

21 minutes

Took

One sitting

Find 20 winnable keywords for salesbuilder.dev (AI agents that get a founder their first customers).

Achieved

20 of 20 keywords

Spent

$0.12 of $1.00

Calls

78

Agent time

6 minutes

Took

One sitting

It plans before it spendsCopy Anchor Link

A goal is a hub Goal with one Key Result, due within 31 days. The agent works through the call’s agentic Function, whose budget is the most it may spend.

It starts from the call’s default program, the one a Workflow runs with every input set. Free counts size each step before anything is bought.

Each phrase of yours is a hard rule it never relaxes. A filter it adds to get there, such as a headcount range standing in for unit count, is soft, and the plan says so.

In try mode the plan comes back with its estimate, and nothing is spent. To run that exact plan, pass it back as plan.

1. Read your words: each phrase becomes a hard rule, and any filter the agent adds is marked soft.
2. Plan from the default program of the Workflow, agent task or Service the sentence matches, sized by free counts.
3. Price each step at its unit price, and add one priced contingency step for the likeliest shortfall.
4. Ask at most one question, and only when a reading would change the price by more than twice or the result by more than half.
5. Return the plan and its estimate for $0. Nothing runs until you call it again with send or do.

It runs inside the budget and changes course on the recordCopy Anchor Link

The budget is a limit on the goal’s Action, checked before every call. When the goal is met, it stops. When the next call fits what is left, it makes it.

When the next call does not fit, it stops and opens an approval with its projection: “43 of 50 found. About 61¢ more finds the last 7. Your budget has 18¢ left.” A person answers it; the agent never does.

Every change of course is a Decision that supersedes the plan, with the options it weighed and its reason. A fallback between routes is the platform’s job and records no Decision.

The receipt shows every call and whyCopy Anchor Link

The receipt is a view over one Action. It opens on the Decisions, then gives one line per call: how many ran, what they cost, and what came back. Every row sits in a fold beneath, and the totals equal the sum of the rows.

Agent minutes and wall-clock days are two figures. A goal that waits on replies runs for weeks on a few minutes of agent time. Ad spend sits beside the budget, never inside it: the site bills it to your own account.

The agent’s own model cost sits on its own line as agentUsd, so you can see what the planning cost apart from the work. It is usually 2¢ to 10¢ a goal.

Five goals, workedCopy Anchor Link

The two goals above each ran one call. Three more ran several. One wrote and indexed answer pages until AI answers named Leaseline. One ran a message test on $100 of ad spend, paused two losers early, and asked for $50 more. You said no.

The third is this page’s Run it: Goal.pursue booked five calls with property managers in 19 days and 95 agent minutes. It built the list, wrote the first touches, answered the replies and briefed each call. On day 6 it changed segment, as a Decision, because the larger companies had not replied.

Each call’s own page carries its goal’s plan, its Decisions and its full receipt. Directory submissions carries one more, where the agent switches routes when an API refuses.

Get Leaseline named in AI answers for the 10 questions small landlords ask most about tracking rent.

Achieved

10 of 10 questions answered by an indexed page

Spent

$1.94 of $15.00

Calls

107

Agent time

60 minutes

Took

21 days

Find out which message gets small landlords to start a trial. Put up to $100 into ads.

Achieved

1 of 1 winning message, likely

Spent

$0.28 of $5.00

Ad spend

$100.00 on your account

Calls

41

Agent time

14 minutes

Took

5 days

The 50-lead goal, call by callCopy Anchor Link

The 50-lead goal, call by call

The receipt of Build a verified list in goal mode. Run for real, it found 50 of 50 verified, and six Decisions explain every change of course.

Decisions

1. D1, rows 1 to 5size four queries for free, then buy 150 rows from the closest one.All real estate matched 5,150 people. Adding the keyword “property management” brought it to 882, the nearest fit to your sentence, so row 5 bought from that query.
2. D2, rows 54 to 57find 4 addresses again, by name and the current company’s domain only.The first find for these 4 people returned an address at another company, such as gwinnettcounty.com for a partner at tmg-atl.com. An address at a past employer bounces or reaches the wrong person.
3. D3, rows 105 to 112 and 136send 9 catch-all and risky results to ContactPoint.verifyEmail({ depth: 'deep' }), at $0.004 each.A catch-all domain accepts every address, so the standard check cannot confirm the mailbox. All 9 came back deliverable.
4. D4, rows 113 to 116find 4 addresses again, leaving out the route that produced each one.The check called them invalid. One came back new and valid; the other 3 stay off the list, each with its reason below.
5. D5, rows 118 to 123re-size four narrower queries for free, then buy the 51 rows of “Leasing Residential Real Estate”.The second page of the first query came back empty. That industry was the most precise match, and small enough to read whole.
6. D6keep 10 addresses at a domain other than the company’s, such as choicenewyork.com for choice-ny.com.Each checked deliverable, and none sat at a past employer on the person’s profile, which is what D2 tests for.

By call

Call	Calls	Cost	Outcomes
Prospect.count	8	$0	8 counted
Prospect.find	3	$0.05738	3 listed
ContactPoint.findEmail	62	$0.172	60 found · 2 none found
ContactPoint.verifyEmail	54	$0.026	4 invalid · 41 deliverable · 7 risky · 2 catch-all
ContactPoint.verifyEmail({ depth: 'deep' })	9	$0.036	9 deliverable
Spent	136	$0.29	of $4.00, $3.71 left

Agent time 21 minutes, in one sitting.

Every call, 136 rows

Loading the rows.

The 20-keyword goal, call by callCopy Anchor Link

The 20-keyword goal, call by call

The receipt of the Keyword opportunity map in goal mode, including the switch to search-box suggestions when related terms had nothing for founder phrasing.

Decisions

1. D1, rows 8 to 16switch from related terms to phrase match and search-box suggestions.Related terms had no reading for founder phrasing such as “how to get first customers”. Phrase match and suggestions returned 205 rows and 52 suggestions for the same words.
2. D2, row 17measure the 63-term shortlist in one batch.Difficulty is priced by the batch of up to 1,000 terms, so 63 terms cost one batch. The 19 terms with no reading cost nothing.
3. D3, row 78read salesbuilder.dev’s own authority as the baseline.At an authority of 1, a term is winnable only where page one already holds weak or user-written pages, which is what the soft slots column counts.

By call

Call	Calls	Cost	Outcomes
Keyword.find	10	$0.004	8 listed · 2 no reading
Keyword.find({ mode: 'suggestions' })	6	$0.0051	6 listed
Keyword.measure	1	$0.022	1 measured
Keyword.search	28	$0.0042	28 read
Domain.benchmark	33	$0.08861	33 read
Spent	78	$0.12	of $1.00, $0.88 left

Agent time 6 minutes, in one sitting.

Every call, 78 rows

Loading the rows.

A goal that works becomes a Workflow, then a PlayCopy Anchor Link

When every change of course on a receipt can be written as a rule, the receipt offers “Make this a Workflow”. Each Decision becomes a declared rule, such as “if fewer than 50 are valid, read 200 more companies, at most twice”.

The draft deploys as your own call, priced before it runs (Compose your own). “Keep doing this every week” is a Play’s job, not a goal’s, so the receipt offers the Play instead.

Four types carry a goal from sentence to receiptCopy Anchor Link

Every field, the two ways in, streaming and failure are on the reference page.