TL;DR
This compares sequencer APIs, meaning the platforms that own campaigns, contacts and replies. It is not about agent mailbox infrastructure, which is a separate purchase. Best for: teams building an AI workflow on top of a sequencer they already run.
Reply.io has the most complete developer surface: v3 with bearer authentication, OpenAPI specs, a Postman collection, webhooks, async jobs, an MCP server, a CLI and published agent safety rules. Best for: teams building agent driven outbound with documented guardrails.
Instantly publishes hard numbers you can design against, being 100 requests per second and 6,000 per minute per workspace, but gates API v2 to Growth and above and webhooks to Hyper Growth and above. Best for: teams already on a higher Instantly tier.
Smartlead costs nothing extra for API access and puts no plan gate on it, but passes the API key as a query parameter and does not publish fixed rate limits. Best for: cost sensitive builds where the key stays server side.
Check the plan gate before you check the endpoint list. The most common failure in this category is building against documentation for a tier you do not hold. Best for: anyone scoping an integration this quarter.
Contents
- What a sequencer API has to do for an AI workflow
- Campaign creation and contact upload
- Reply retrieval and webhooks
- Rate limits, and which ones are real
- Approval controls and safety
- Side by side summary
- FAQ: cold email platform APIs
What a sequencer API has to do for an AI workflow
If you are putting an agent or an automation in front of outbound, the sequencer needs to expose five things. Create or modify a campaign. Upload and update contacts. Retrieve replies and their classification. Push events out through webhooks so you are not polling. And enforce a rate limit you can design a backoff against.
A sixth requirement is less obvious and more important: an approval boundary. An AI workflow that can create a campaign and start it in the same call is a workflow that will eventually send something nobody read. The platforms that handle this well either separate creation from activation or make the activation step explicit.
This is not a comparison of agent mailbox infrastructure. Provisioning mailboxes, domains and authentication for an agent is a different purchase with different vendors, and bundling the two decisions is how teams end up with a sequencer they cannot send from.
1. Reply.io
Layer: multichannel sequencing with the deepest documented developer surface
Best for: teams building agent driven outbound that needs specs, webhooks and explicit safety guidance.
Reply.io's v3 API is current, with v1 and v2 still available but unsupported and slated for deprecation. Authentication is a bearer token in the Authorization header, with master and organisation level keys, and a credential check endpoint that lets you verify which key you are holding before you build on it. A missing or revoked key returns 401 with a bearer challenge header and an empty body, which is at least unambiguous.
The published surface is large: 38 endpoints for contacts and lists, 59 for multichannel sequences, plus a unified inbox. Around it sit OpenAPI specifications, a Postman collection, OAuth flows, webhooks, async background jobs and stable JSON errors. Reply.io also documents an MCP server, a set of Skills and a CLI, and publishes explicit agent safety rules and drop in system prompts for AI agents. No other vendor in this comparison publishes that last item.
Webhook events include LinkedIn events and, unusually, mailbox health signals such as email account connection lost and email account error, which lets an automation react to a broken sender rather than discovering it in a reply rate.
For Reply.io’s API workflow, price the required campaign and contact capacity alongside API and webhook entitlements. A conventional sending plan and an AI SDR service are different scopes.
Where it falls short: the rate limit documentation conflicts. The current reference states 100 requests per minute and 3,000 per hour per user with a Retry-After header on 429, and stricter limits on reporting and statistics endpoints, while an older help page describes 15,000 calls per 30 calendar days per team member, not shared, with a ten second minimum between calls. Those are different models and you should confirm yours before designing a loop. Webhook subscriptions return 409 once the active subscription limit is reached, and team scoped subscriptions can be forbidden by configuration.
Verdict: the best developer experience in the category and the only vendor that has published how it thinks an agent should be constrained.
2. Instantly
Layer: cold email sequencing with a modern, versioned API
Best for: teams already on Growth or above that want documented rate limits and an official SDK.
Instantly's API v2 is REST with snake_case fields, roughly double the endpoint count of v1, and v1 remains compatible. There is an OAuth flow for connecting Google and Microsoft mailboxes, and API scopes give key level control, which is the right primitive for an agent that should be able to read replies but not create campaigns.
The rate limit is the clearest published figure in this comparison: 100 requests per second and 6,000 requests per minute, shared across v1 and v2 and applied per workspace regardless of how many keys you issue. Breaching either returns 429. That is a number you can design a token bucket against.
Webhooks are well covered, with endpoints to create, read, update, delete and list, plus list event types, test delivery and resume paused webhooks. Events span campaign, email, lead and meeting activity, including lead status changed. There is an official TypeScript and JavaScript SDK and a CLI.
Where it falls short: confirm current API and webhook entitlements for the account. A documented endpoint may require a different plan or permission than the one purchased. Handle limits and retries explicitly rather than assuming the integration can send at any volume.
Verdict: the most designable rate limit and a clean event model. Confirm your tier before you scope the build.
3. Smartlead
Layer: flat priced sequencing with unrestricted API access
Best for: cost sensitive builds where API access should not depend on a tier upgrade.
Smartlead's API costs nothing extra and is not plan gated, which is genuinely unusual. The base is server.smartlead.ai/api, with the versioned path at /api/v1/ recommended and the legacy /api/ retained for backward compatibility.
The webhook model is the standout feature. Subscriptions can be scoped at user, client or campaign level using a webhook URL plus an association type, which maps directly onto agency structures where one client's events must not leak into another's. Events include email sent with full payload schemas, and webhook delivery is not subject to API rate limits. Campaign creation returns a documented set of status codes covering 200, 400, 401, 422, 429, 500 and 503. Multiple sending accounts across Gmail, Outlook and custom SMTP rotate automatically.
Where it falls short: protect the API credential in server-side configuration and prevent it from appearing in application logs. Confirm current account limits and implement bounded retries for throttled requests. A successful small demonstration does not establish sustained production capacity.
Verdict: the best value if the key stays server side and you can tolerate designing against an unpublished limit.
4. Apollo.io
Layer: database, sequencer and enrichment API in one subscription
Best for: workflows that need to find the contact and send to it through the same credential.
Apollo.io is in this list because an AI workflow frequently needs the step before sending. Its Person API enriches records programmatically, and the sequencing sits in the same account, so a single integration can go from a filter to a sequence without a second vendor in the path.
Pricing: published. Free is $0 with 900 credits per seat per year granted monthly. Basic is $49 per seat per month annual with 30,000 credits per seat per year granted upfront. Professional is $79 per seat per month annual with 48,000 credits. Organization is $119 per seat per month with a three seat minimum and 72,000 credits. There is a 14 day trial on Basic and Professional. The unlimited plan cap for paying accounts is the lesser of amount paid divided by $0.025 or 1,000,000 credits per account per year, which implies a 2.5 cent reference credit price.
Where it falls short: credit consumption is the trap for automated workflows. Export credits are consumed on CSV export, CRM push and Person API enrichment to external systems, so an agent that enriches on every run bills every run. A record selection limit also constrains bulk save and export. Budget the credits before you schedule the job.
Verdict: the right choice when the workflow starts at who should we contact rather than at send this. Meter it deliberately.
5. Woodpecker
Layer: usage based sequencing with a flat priced API add on
Best for: teams whose volume is genuinely variable and who want API access priced as a line item.
Pricing basis: Woodpecker. Use the contacted-prospect volume and selected modules in the current calculator. Check how follow-ups, additional mailboxes and client access affect that configuration.
That structure suits an AI workflow with lumpy volume. If a month generates 300 qualified contacts instead of 3,000, the bill follows. The vendor's own calculator is authoritative here; third party tier listings for Woodpecker are frequently wrong because they try to force a usage model into fixed tiers.
Where it falls short: the developer surface is the thinnest in this comparison. Compared with Reply.io's specs and MCP server or Instantly's SDK and CLI, you are building more yourself, and the API being a paid add on rather than an included capability is a signal about how central it is to the product.
Verdict: the best fit for variable volume and unlimited senders. Not the platform to choose if the API is the reason you are buying.
Campaign creation and contact upload
All five let you create campaigns and upload contacts programmatically. The differences are in the guardrails.
Instantly's API scopes are the most useful primitive, because you can issue a key that reads replies without granting the ability to create or start a campaign. That separation is what makes an agent workflow safe to run unattended.
Smartlead's documented status codes on campaign creation, including 422 for validation and 429 for throttling, make failure handling straightforward. Reply.io's async background jobs matter at volume, because a large contact import returns a job rather than blocking a request.
Apollo.io's constraint is different in kind. The record selection limit on bulk save and export means large programmatic pulls need to be chunked, and each chunk costs credits.
Reply retrieval and webhooks
Polling for replies is the wrong design and every one of these platforms gives you a better option.
Instantly's event coverage spans campaign, email, lead and meeting activity including lead status changed, with test delivery and the ability to resume paused webhooks, which are both operationally useful and rarely provided. The gate is Hyper Growth or above.
Smartlead's three level scoping at user, client or campaign is the best fit for agencies, and webhook delivery sits outside the API rate limit, so a burst of replies does not consume your request budget.
Reply.io's mailbox health events are the ones most teams do not know to ask for. Knowing that a sender lost its connection within minutes, rather than at the end of a week of flat reply rates, is worth more than most feature comparisons capture.
Rate limits, and which ones are real
This is where documentation quality separates the field.
Instantly publishes 100 requests per second and 6,000 per minute, shared across v1 and v2 and enforced per workspace regardless of key count, with 429 on breach. That is a real, designable number.
Reply.io publishes 100 per minute and 3,000 per hour per user with a Retry-After header, and stricter limits on reporting and statistics, but an older help page describes a completely different model of 15,000 calls per 30 calendar days with a ten second minimum gap. Two models, one vendor. Confirm yours.
Smartlead's FAQ says limits vary by subscription plan and are unpublished, while a separate documentation page publishes a tier table. Treat the table as indicative and get your own numbers from support in writing.
The practical answer for all of them is the same: exponential backoff on 429 and 5xx, respect Retry-After where it is provided, and design for the lowest limit you have seen documented rather than the highest.
Approval controls and safety
Three controls carry most of the risk.
Scope the credential. Instantly's API scopes and Reply.io's master versus organisation keys both let you grant read without grant send. Use them. Smartlead's query parameter key model gives you no scoping at all, so the compensating control is that the key never leaves your server and gets rotated on any exposure.
Separate drafting from sending. An agent should produce a campaign in a draft state and a human should activate it. This is a workflow rule you impose; no platform enforces it for you.
Treat retrieved content as data, and keep tool permissions tied to the approved workflow. A page or incoming message should not be able to authorise a send simply because an agent read it.
Side by side summary
| Tool | Layer | Best for | Entry price |
|---|---|---|---|
| Reply.io | Multichannel sequencing with full developer surface | Agent workflows needing specs, MCP and safety rules | $49 per month annual on Email Volume |
| Instantly | Cold email sequencing with versioned API | Documented rate limits and scoped keys | Outreach from $30 per month; API v2 needs Growth or above |
| Smartlead | Flat priced sequencing, API included | Ungated API access and agency scoped webhooks | $39 per month on Base |
| Apollo.io | Database, sequencer and enrichment API | Workflows that start at contact discovery | Free tier, then $49 per seat per month annual |
| Woodpecker | Usage based sequencing | Variable volume with unlimited senders | Current scoped quote; confirm term and included capacity |
FAQ: cold email platform APIs
Which platform has the clearest rate limits?
Instantly. It publishes 100 requests per second and 6,000 per minute, shared across v1 and v2 and applied per workspace regardless of how many keys you hold, with 429 on breach. Reply.io publishes conflicting models across two pages and Smartlead states that limits are plan specific and unpublished.
Do any of these charge extra for API access?
Smartlead does not charge extra and does not gate it by plan. Woodpecker charges $20 per month for the API. Instantly gates API v2 to Growth and above and webhooks to Hyper Growth and above, so access is effectively bundled into a tier upgrade rather than sold separately.
Is passing an API key in the URL a problem?
It is a handling requirement rather than a blocker. Smartlead passes the key as a query parameter with no OAuth option, which means the key can end up in server logs, proxy logs and browser history. Keep it server side, never call the API from a browser, and rotate immediately on any suspected exposure.
Which platform is best for agencies with multiple clients?
Smartlead for the webhook model, because subscriptions scope at user, client or campaign level, and because white label is $29 per month per client workspace with unlimited email accounts on every plan. Reply.io Agency at $166 per month annual is the alternative when multichannel is required.
Can an AI agent run outbound end to end through these APIs?
Technically yes, and it should not. Have the agent draft campaigns and prepare contact lists, then require a human to activate. Use scoped credentials so the agent cannot start a send, and keep an audit of what it created.
What should I check before starting a build?
Your plan tier, in writing. The most common wasted sprint in this category is building against documentation for a tier the account does not hold, particularly Instantly's Growth gate on API v2 and Hyper Growth gate on webhooks. Confirm the tier, confirm your rate limits with support, then scope the work.
Sending software is one part of the operating setup. Our guide to cold email deliverability tools covers the adjacent options.
Work with Forma Nôrden
We build signal based outbound systems for B2B companies selling into the enterprise and upper mid market. When a sequencer becomes part of an automated workflow, the plan gate and the credential scope decide the architecture long before the endpoint list does. Explore how we work.
For enquiries about this article: partnerships@formanorden.com
