Credits

Credits are a single org-wide pool, shared by everyone in your org — they aren’t split up per seat or per person. Every trigger run and chat conversation draws from the same balance.

​

Two pricing versions

In October 2026, Swan started rolling out a new way to use credits. Every org is on one of two pricing versions: Both versions run side by side while Swan gradually moves orgs from v1 to v2. Orgs created from October 12, 2026 start on v2 by default; the gradual move applies to orgs created before then. Your org uses one version at a time, and every charge follows that version. Swan knows which version your org is on — ask it, or open Settings → Subscription, which shows the plans for your version. v2 was built to make credits easier to predict and control:

• Clearer. On v1, many small actions each had their own price, so it was hard to know what cost credits. On v2, data has a public price list, and everything else is the work Swan does.
• No penalty for getting work done. On v1, every CRM update, Slack message, or sequence build cost a credit, which discouraged automations that act in your other systems. On v2, those actions have no price of their own.
• Closer to the real work. Credits follow how much work a task really takes, so Swan’s incentives and yours line up: simpler, focused work costs less.
• More control. You decide how complex and how long-running your tasks and automations are, and where it’s available, which model does the work.

v1 credits and v2 credits are different units. Don’t compare a v1 balance with a v2 balance credit for credit. The exact prices for both versions are in credit costs.

​

What you pay per credit

The price of one credit depends on your plan. Standard plans have a fixed price per credit. Custom and enterprise plans are priced individually. Your plan and its credits are on Settings → Subscription. Custom and annual plans can differ — contact the Swan team for those.

​

What costs credits on v1

Most of Swan’s paid actions cost 1 credit each: web research, scraping a URL, building a sequence, generating a presentation, writing to your CRM, running code, and calling a custom tool (including tools from an MCP server you’ve connected). A few actions cost more because they call more expensive third-party data providers:

• Enriching a full contact or company profile costs 1 credit, charged only when enrichment data is returned.
• Enriching an email address or phone number costs 2 credits each, charged only on a successful match.
• Fetching business/intent events costs 1 credit per event returned.
• Fetching latest website changes costs 1 credit per new provider date for each company. Repeated same-date results, unmatched companies, and no-data companies are free.
• Finding similar companies costs 1 credit per 5 results returned, rounded up; no matches cost nothing.

Costs for tools from other connected apps (HubSpot, Salesforce, and so on) vary by the specific action — most simple writes cost 1 credit, but check the tool’s own description for anything unusual. Most trigger sources also carry a processing fee, separate from whatever the run does — this ranges from 1 to 2 credits depending on the trigger type (for example, a scheduled or LinkedIn-engagement trigger costs 1 credit to fire; a business-event or intent-signal trigger costs 2). Website Changes costs 1 credit when a new provider-reported change is classified, even when the filter rejects it; unchanged, missing-data, and excluded-company checks are free. On v1, Swan’s AI itself is free. You pay only for the actions it takes.

​

What costs credits on v2

On v2, credits come in two kinds. Both come from the same balance. Data credits. Data lookups — enriching a contact or company, finding an email or phone number, web research, scraping a page, fetching business events, finding similar companies — each have a fixed price per result. The full list is public in credit costs, and Swan sees each charge as it happens. A lookup with no result costs nothing. Swan prices its data competitively against what the same data costs elsewhere on the market. Triggers that bring in data have a data-credit fee too: business events, Bombora intent, website visitors, and Website Changes. Other trigger types have no fee on v2. Intelligence credits. Everything Swan does to think and work — reading your data, reasoning, researching, writing, and deciding which actions to take — uses intelligence credits. The amount depends on:

• How complex the task is. A quick question uses fewer than a multi-step plan.
• How much context Swan works with. Long conversations, large files, and big lists use more.
• How long the work runs. More steps use more. Each sub-agent is its own run and uses its own intelligence credits.
• Which AI model does the work. A lighter model uses fewer intelligence credits for simple work. Some orgs can already choose the model in the app for some conversations and automations. This isn’t available to everyone yet, and it will reach more places over time.

Actions that don’t bring in data — CRM writes, building sequences, running code, Slack messages, LinkedIn reads, and tools from MCP servers and connected apps — have no price of their own on v2. They are part of the work that intelligence credits cover.

​

Spending fewer intelligence credits

• Keep tasks focused. Start a new conversation for a new topic instead of continuing a very long one.
• Filter a list before you ask Swan to work through it record by record.
• For the same simple action on many records, let Swan batch it in one code run instead of one sub-agent per record.
• If your org can choose the model in the app, use a lighter model for simple, repetitive work.
• Keep automations short and specific. An automation that fires often multiplies whatever each run uses.

​

Batch discounting

On v1, if a single run performs the same non-data action many times — for example, updating ten CRM records with the same tool — you’re charged once for that batch, not once per record. Actions that pull in new external data per record (like enriching a contact, company, email, or phone number, or fetching a URL) are the exception: those are charged per record, since each one is a separate lookup against a paid data provider. On v2, non-data actions have no price of their own, so there’s nothing to coalesce. Data lookups are still charged per result. Batching still saves credits on v2: one code run that does 50 actions uses far fewer intelligence credits than 50 separate steps.

​

When credits are charged

Before a new run starts, Swan checks that your org has at least 10 credits remaining. Long-running trigger executions can span multiple work slices; after they start, later slices only require one remaining credit. Credits are applied as execution work is persisted, so a run that reaches zero can be stopped at its next slice boundary. These rules are the same on both versions. A trigger fee is charged once, when the run starts. A run that fails or is cancelled later still pays its trigger fee. Onboarding conversations use credits like any other chat, on both versions. Before October 2026, they were free. On v2, a single step can use a fraction of a credit. Swan keeps the exact amount and shows whole credits in the app, rounded to the nearest credit.

​

Rollover

On monthly plans, unused credits carry over into your next billing cycle, up to a cap of twice your plan’s monthly allotment — anything beyond that cap doesn’t carry forward. On annual plans the cap is lifted: unused credits keep rolling forward without limit within your 12-month term.

​

Usage alerts

Your org gets a notification when you cross 50% and 90% of your monthly credit allotment, and again once you’ve used it all. Check current usage any time at Settings → Usage.

​

Running out of credits

Once your balance drops below 10, Swan blocks new chat runs and new trigger-driven runs — you’ll see an insufficient-credits message instead. Delegating work requires at least 10 credits overall and five available credits per delegated task, whichever is greater. This is a balance check, not a charge. Triggers themselves are not paused or disabled; they simply fail to start until enough credits are available again. Your dashboard, history, and existing data all stay fully accessible either way. If you need more credits before your next renewal, you can buy a one-time top-up from Settings → Subscription — see plans and billing.

​

Common questions

What changed in October 2026? Swan started rolling out v2. On v2, you pay data credits for data lookups, at fixed public prices, and intelligence credits for Swan’s work, based on how complex and how long that work is. Actions such as CRM writes and sequence builds no longer have their own price, and most trigger types have no fee. Orgs on v1 keep per-action credits until they move to v2. Onboarding conversations also started to use credits. Why is v2 better? It’s easier to predict and control. On v1, it was hard to tell which of many small actions cost credits, and every action in your other systems — a CRM update, a Slack message, a sequence build — cost a credit, which discouraged automations that get real work done. On v2, data has a public price list, actions have no price of their own, and credits follow how much work a task really takes. That lines up Swan’s incentives with yours, and you control the cost through how complex and how long your tasks are and, where available, which model does them. Which pricing version is my org on? Ask Swan — it knows your org’s version. Settings → Subscription also shows the plans for your version. What are data credits and intelligence credits? They are the two kinds of v2 charges, from one balance. Data credits pay for data lookups at fixed public prices. Intelligence credits pay for Swan’s work — reasoning, research, writing — and depend on how complex the task is, how much context it needs, how long it runs, and which model does it. Why did a chat cost credits on v2 when Swan didn’t use any paid tool? On v2, Swan’s work uses intelligence credits. Reading your data, reasoning, and writing a reply all count. On v1, the same chat costs nothing unless Swan takes a paid action. How much does one credit cost? It depends on your plan. Standard plans have a fixed price per credit; custom and enterprise plans are priced individually. See Settings → Subscription. Does Swan tell me how many credits a run will use before I start it? No, there’s no upfront estimate — credits are deducted based on what the run actually does. You can use the cost breakdown above to estimate roughly, and check usage after the fact in your execution history. For large jobs, Swan gives an estimate and asks before it starts. On v2, Swan can give exact data-credit prices, but intelligence credits are an estimate, because they depend on how the work goes. If I run a batch of 50 identical CRM updates, am I charged 50 credits? No. On v1, identical, non-data actions in the same run are coalesced into a single charge. On v2, CRM updates have no price of their own — they are part of the run’s intelligence credits. Per-record data lookups (enrichment, business events, URL scrapes) are still charged per record on both versions. What happens if a run goes over my remaining balance mid-way? It can finish its current work slice. If it reaches zero and needs another slice, Swan stops it before more work begins. Are trigger runs paused automatically when I run out of credits? No. The trigger stays enabled, but each firing will fail with an insufficient-credits message until you have credits again. Do unused credits roll over indefinitely? On monthly plans, no — rollover is capped at twice your plan’s monthly credit allotment. On annual plans, credits roll over without a cap for the duration of your term. Can I see which user or trigger is using the most credits? Yes, credit usage is broken down by execution at Settings → Usage, which you can use to see which triggers or users are the heaviest consumers. On v2, the page lists every automation and teammate with their spend split into intelligence, data and signal fees for the billing cycle or any date range, plus a typical and a high cost per run and the costliest runs. See analytics for the page itself.