Back to vendors
S

Stepper

Also known as: Stepper.io, stepper.io, Stepper by Paperform

Visit site
Entry pricePro $19 a monthFull pricing detail

Stepper, from Paperform, builds workflow automations from a plain language description, combining a conversational AI builder with a visual editor, reusable components and triggers and actions across 489 apps. Skill Sets bundle those authenticated actions into scoped toolkits that Claude, Cursor, any MCP client or the stpr command line tool can call, with locked parameters and a log of every call.

Stepper is Paperform's AI native workflow automation platform, pitched to small businesses and solo operators as a Zapier or Make alternative with a flat price. Accounts are created through Paperform, and the product runs as a hosted service.

A conversational assistant builds a workflow from a description, explains it and changes its logic on request, while the visual editor stays available for manual edits. Workflows start from webhooks, polling, schedules or app events, run conditional branches, loops, error handling and JavaScript steps, and act across 489 apps, from Gmail, Slack and Google Sheets to HubSpot, Salesforce, Stripe, Shopify and Jira, plus HTTP requests to any URL.

AI steps answer questions, decide routes, generate images, read PDFs and images with OCR and run web searches, on Stepper's built in providers or the customer's own OpenAI, Anthropic, Google, Mistral or other keys. Components package repeated steps into a library that many workflows share. Skill Sets turn chosen actions and components into an MCP endpoint and a CLI toolkit for outside agents, with parameters that can be locked, per role tokens using static bearer tokens or OAuth with PKCE, and a 30 day log of every call.

The Free plan includes unlimited workflows, 200 steps a month and 5,000 credits; Pro costs $19 a month for unlimited steps under a fair use limit of about 250,000 a month and 15,000 more credits, with extra credits at $1 per 1,000 that never expire. Credits pay for AI steps, email, SMS and premium APIs, and bringing your own keys reduces that spend; a Team plan with collaboration and advanced permissions has not launched. Stepper is GDPR compliant and holds no SOC 2, ISO 27001 or HIPAA certification.

Stepper fits small teams that want to describe automations instead of mapping every field, keep costs flat as runs grow, and hand a controlled set of business actions to Claude, Cursor or their own agents. It fits less well for teams that need agents that plan their own steps, a built in knowledge base or memory, human approval steps, single sign on and audited compliance, or self hosting.

Vendor details

Canonical URL

https://stepper.io

Category

Agent builder

Subcategory

Conversational workflow builder across 489 apps with reusable components, and Skill Sets that expose scoped actions to AI agents over MCP and a CLI

Funding status

Product of Paperform, a self funded and profitable software company.

Company status

first party product

Use cases & customers

Primary use cases

AI lead qualification and CRM updatesinvoice and receipt extraction with OCRdrafting replies to new emailsapproval and onboarding workflowscontent distribution and change notificationsscoped tool access for AI agents through MCP

Target customers

small businesses and solo operators automating admin and sales workteams moving off Zapier or Make to avoid per task billsPaperform users automating form and signature workflowsdevelopers giving Claude, Cursor or their own agents scoped access to business appsagencies running client automations on a flat budget

Deployment options

Hosted SaaS

Integrations

489 apps with triggers and actions, including Gmail, Microsoft Outlook, Slack, Microsoft Teams, Google Sheets, Google Calendar, Notion, Airtable, HubSpot, Salesforce, Pipedrive, Stripe, Shopify, Xero, Jira, Linear, GitHub, Zendesk, Intercom, Twilio, Typeform, Calendly and Paperform, plus an HTTP Request action, webhooks, custom connectors and JavaScript code steps. Model providers include OpenAI, Anthropic, Google Gemini, Mistral, DeepSeek, xAI, Cohere, Perplexity and OpenRouter. Skill Sets expose chosen actions over MCP to Claude, Cursor, Windsurf and other clients, and through the stpr CLI on npm.

In practice

A consultancy describes its intake process to the assistant, which builds a workflow that reads each new Paperform submission, has an AI step score the lead, adds a HubSpot contact and posts qualified leads to Slack.

A bookkeeper runs the invoice template so that PDF invoices are read with AI OCR and their supplier, amount and due date are written to a Google Sheets tracker.

An engineering lead creates a Skill Set that lets Claude read a spreadsheet, create Jira tickets and post only to the #alerts Slack channel, issues a token for it, and checks the call log to see each action the assistant took.

Agentic Index coverage score

7.0 / 14 capabilities · 50%

Integrations & Tool Calling Full

Workflows act across 489 apps, each with its own triggers and actions, among them Gmail, Slack, Google Sheets, HubSpot, Salesforce, Notion, Airtable, Stripe, Shopify, Jira, Zendesk, Xero, Microsoft Teams and Paperform itself, where actions reach as far as updating form fields, products and submissions.

An HTTP Request action calls any URL with full control of method, headers and body, custom connectors cover systems outside the catalog, and teams can bring their own credentials. Skill Sets hand the same authenticated actions to outside AI agents, so an assistant can add a spreadsheet row, post to Slack or create a Jira ticket through one connection.

Sourcestepper.ioread 2026-10-18

Workflow Orchestration Full

A conversational builder turns a plain language description into workflow steps and edits existing logic on request, and the visual editor shows the result for manual changes.

Workflows support conditional branches, loops and iterators, error handling and JavaScript code steps, and components package a reusable set of steps into a library that any workflow can call, so a lookup or authentication routine is built once and changed in one place.

The steps themselves are fixed when the workflow is saved: an AI step can classify a record or pick a route inside the flow, but no model plans the sequence at run time.

Sourcestepper.io/automation/stepper-vs-zapierread 2026-10-18

Knowledge Grounding & RAG Partial

The built in AI app answers questions about or extracts text from images and PDFs, runs web searches, and takes whatever data earlier steps pass into its instructions, so an AI step can work from an invoice, a form submission or a search result. There is no knowledge base of Stepper's own to upload documents into, index and retrieve from with citations. Retrieval over a team's documents means wiring in a separate service from the catalog, such as Pinecone, Dify Knowledge Base, Voyage AI or Jina AI, each on its own account.

Sourcestepper.io/apps/airead 2026-10-18

Human Oversight & Guardrails Partial

Skill Sets put guardrails on what an outside agent can do: each one lists the connected apps and the exact actions allowed, can lock parameter values so that an agent allowed to send a Slack message can post only to #alerts, and issues separate tokens per role that can be revoked at once from the interface. Marketing can get email and CRM while engineering gets project tools and monitoring. Inside workflows, approvals appear as a use case, but no step pauses a run for a person to approve, reject or edit before it continues.

Sourcestepper.io/skill-setsread 2026-10-18

Security, Identity & Governance Partial

Skill Sets authenticate with static bearer tokens for scripts or OAuth 2.0 with PKCE for richer clients, scope each token to chosen apps and actions, and show which token or OAuth client made every call. Accounts sign in through Paperform. Stepper is GDPR compliant and holds no SOC 2, ISO 27001 or HIPAA certification, and single sign on and team roles are absent today, with collaboration features and advanced permissions slated for a Team plan that has not launched.

Sourcestepper.io/skill-setsread 2026-10-18

Observability & Auditability Partial

Every Skill Set call, successful or failed, is logged with the tool name, status, authentication method, timestamp and the token or OAuth client that made it, kept for 30 days and visible in the Stepper interface, so a team can see each action an outside agent took. Workflows themselves have no step level run trace, log export or link to an outside monitoring tool, and nothing records the reasoning behind an AI step's output beyond the result it passes to the next step.

Sourcestepper.io/skill-setsread 2026-10-18

Memory & State Persistence Not documented

Data moves from step to step within a run and lands in whatever records the workflow writes, such as a Google Sheets tracker or a CRM entry, but Stepper keeps no memory of past runs or conversations for its AI steps to draw on. Memory services in the catalog, including Mem0, Letta and Mem, can be called as ordinary actions on their own accounts. That puts the memory in another vendor's store, so Stepper manages no state of its own for its workflows or for the agents that call Skill Sets.

Sourcestepper.io/apps/airead 2026-10-18

Deployment & Data Residency Not documented

Stepper runs only as a hosted service, with no self hosted edition, no open source release and no private cloud option, so workflows, credentials and logs live in Stepper's cloud; self hosting is slated as coming soon. Bringing your own API keys keeps AI and email calls on the customer's own provider accounts, which changes who bills for those calls but leaves the platform where it is, and the Skill Sets MCP endpoint is likewise a single hosted address at mcp.stepper.io. No region choice is offered.

Sourcestepper.io/automation/stepper-vs-zapierread 2026-10-18

Prebuilt Agents / Templates / Packs Partial

Live templates can be copied and run straight away, such as one that extracts invoice details from PDF invoices with AI OCR and one that drafts replies to new Gmail messages while skipping notifications and transactional senders. Component libraries let a team assemble its own reusable building blocks, and the same components can be exposed to agents as single tools. The template set is small next to the 489 app catalog, and there are no packaged agents for a role or business function.

Sourcestepper.ioread 2026-10-18

Triggers & Channel Coverage Full

Workflows start from instant and webhook triggers, polling triggers and scheduled runs, and from app events across the catalog, such as a new Paperform submission, a partial submission, a Papersign event, a new Gmail message or a new message in a chat channel. Components run when another workflow calls them, and through Skill Sets an outside agent, a terminal script or any MCP client can call an action or a whole component on demand and poll for its result. Stepper has no chat or voice channel of its own; conversations reach a workflow through the apps that trigger it.

Sourcestepper.io/automation/stepper-vs-zapierread 2026-10-18

Model Flexibility & Routing Full

AI steps run on Stepper's built in providers by default or on the customer's own keys, with OpenAI, Anthropic, Google Gemini, Mistral, DeepSeek, xAI, Cohere and Perplexity in the catalog alongside OpenRouter for reaching other makers. Each AI step names its own provider, so one workflow can read a document with one maker's model and draft a reply with another's, and bringing your own key moves the cost of those calls off Stepper credits onto the provider's bill. The builder makes that choice step by step, and there is no automatic routing or fallback between providers.

Sourcestepper.io/apps/airead 2026-10-18

APIs / SDKs / MCP Extensibility Partial

Each Skill Set is served as a standard MCP endpoint at mcp.stepper.io for Claude, Cursor, Windsurf or any MCP client, with built in meta tools that handle dynamic parameters, dropdown options and async polling so an agent can fill in a form step by step.

The stpr command line tool on npm discovers, inspects and calls the same actions from a terminal with OAuth login, saved profiles or a static token, and multistep workflows from a component library become single callable tools that run asynchronously. Most of those tools are actions in connected apps such as Slack or Jira.

There is no REST API or SDK for creating, editing or managing Stepper workflows themselves, so the MCP endpoint and the CLI are the only programmatic way in.

Sourcestepper.io/skill-setsread 2026-10-18

Testing, Debugging & Optimization Not documented

The conversational assistant can explain what an existing workflow does, and the stpr tool returns an action's live parameter schema without executing it unless the call flag is set, which lets a builder check inputs before an agent uses them. Beyond that there is no test run against sample data, replay of a past execution, version history or evaluation of AI step outputs, so checking a workflow or a Skill Set before it goes live comes down to running it and reading the result.

Sourcenpmjs.com/package/stprread 2026-10-18

Browser / Computer-use Not documented

Stepper does not run a browser or a desktop of its own, and its AI app reaches the web only through searches.

The catalog does carry browser services, among them Browserbase, Browserless, Hyperbrowser, Steel, Browse AI and Firecrawl, each on its own account and plan, so a workflow can send page work to one of them as an ordinary action and use the result in the next step.

Everything about how those sessions behave lives with the other company: Stepper neither shows the session nor records what happened on screen, and a team that needs a portal filled in picks, pays for and governs that browser engine itself.

Sourcestepper.io/apps/airead 2026-10-18

The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded

Pricing

Pro $19 a month

Flat monthly plan for steps, with prepaid credits that never expire for AI steps, email, SMS and premium APIs

Free tier

Included quota

Free: 200 steps a month and 5,000 credits. Pro: unlimited steps under fair use (custom plan offered above about 250,000 steps a month) and an additional 15,000 credits. AI help while building draws on a free daily credit quota.

What is public

Plan prices, step allowances, included credits, the credit top up rate and the fair use threshold.

Billing mechanics

Self serve monthly subscription through a Paperform account, with prepaid credits bought from the dashboard.

Cost watchouts

AI steps, email and SMS spend credits unless own keys are connected, and workflows pause when credits run out.

Variable cost rationale

Steps are unlimited on Pro for a flat $19, and only AI, email, SMS and premium API actions draw on prepaid credits, which bringing your own API keys can avoid; heavy AI use without own keys is the main variable.

Additional watchouts

Running out of credits pauses AI dependent workflows until more are bought.

Overage / add-ons

Pro buys extra credits at $1 per 1,000 from the dashboard; credits do not expire or reset, and workflows that depend on credits pause when they run out. Free plans are capped at 200 steps a month.

Sales call required

No, self serve available

Free / trial

Free plan: unlimited workflows, 200 steps a month, 5,000 credits; no credit card required

Lowest paid plan

Pro at $19 a month: unlimited workflows, unlimited steps under fair use, an additional 15,000 credits, extra credits at $1 per 1,000

Commercial notes

A Team plan with unlimited credits, collaboration and advanced permissions is listed as coming soon; accounts far above fair use are moved to a custom plan.

Key ambiguities

How many credits an AI step, email or SMS costs is not stated, so the reach of 15,000 credits is unclear, and whether Skill Sets need the Pro plan is not said.

Cancellation / refund

Cancel any time; workflows run to the end of the billing period; credits are not refundable.

Missing data

Credit cost per action type, annual pricing, Team plan price and Skill Sets plan limits.

Agentic Index verified 2026-10-18

Contact us

Found a vendor we missed? Have feedback on the index? We'd love to hear from you.