A no-code platform for building AI agents and multi-agent Workforces that research accounts, qualify leads and update the CRM, billed by the action and the model call.
What is
Relevance AI
A rep asks for a pre-call brief on Monday, and by Thursday somebody has copied the same five tabs into the same doc for the fourth time that week. Relevance AI is a no-code platform for building AI agents that take that job off the list and run it on a schedule or a trigger. The platform is built from three layers that stack. A Tool is a step chain you build visually from API calls, LLM prompts and Python. An Agent gets a prompt, a model and a set of tools, then decides which of them to call. A Workforce puts several specialists on a canvas and routes handoffs between them. Around that sits a catalogue of roughly 2,000 app connections, built natively for the ones Relevance cares about and extended through Pipedream for the rest. Over a thousand of those app events can start an agent, and an MCP server lets Claude Desktop, Cursor or ChatGPT drive the agents you built from outside the platform. Pay for Relevance AI when a business team wants to build and run its own agents with no engineer nearby, because the visual builder and the evals are aimed at that person. We build that shape for clients with n8n and Claude, where the model bill goes straight to Anthropic and nothing is metered per tool call.
Integrations and the modern GTM stack
Relevance AI splits its catalogue into integrations it built itself and a much longer Pipedream tail, and the difference matters more than the headline number. Native ones arrive with ready-made tool steps and trigger events, while the Pipedream ones give you authentication and an API surface you still have to wire up.
Native GTM connectors | HubSpot, Salesforce, Gong, Apollo, ZoomInfo, Lusha, Marketo and Firmable are built by Relevance, so an agent gets prebuilt actions for reading and writing records rather than a raw API call you configure yourself. These are the ones to start with when you map a GTM workflow onto agents. |
Pipedream catalogue | The jump from a few dozen native apps to roughly 2,000 comes through a Pipedream connection that handles authentication and API access for everything else. There are no ready-made tool steps or trigger events on this tier, so budget build time for anything outside the native list. |
Triggers | More than a thousand app events can start an agent, alongside scheduled runs, custom webhooks, direct HTTP requests and one Tool triggering another. Gating is real here, since WhatsApp, LinkedIn and Telegram triggers need Pro, and Salesforce, Snowflake and Zendesk triggers are Enterprise only even though the Salesforce integration itself is not. |
LLM providers | OpenAI, Anthropic, Gemini, Azure and OpenRouter connect from the Integrations and API Keys page, with Anthropic also reachable through AWS Bedrock or Google Cloud. Adding your own key on a paid plan bypasses Vendor Credits, which puts the model bill back with the provider and takes Relevance out of that part of the cost. |
MCP in both directions | A hosted server at mcp.relevanceai.com exposes your agents, tools and knowledge to Claude Desktop, Claude on the web, ChatGPT in developer mode, Cursor and VS Code, with a dedicated Claude Code plugin the vendor recommends over a manual connection. An MCP client also attaches external servers to an agent. |
Comms, calendar and meetings | Gmail, Outlook, Slack, Microsoft Teams, Zoom, Google Calendar, Twilio, WhatsApp Business and Fireflies cover the surfaces where agents actually deliver work. Calling and meeting agents that use these sit behind the Team plan, so an individual on Pro can send the message but cannot join the call. |
Custom APIs and the platform API | Any Tool can include an API step that hits a REST endpoint with your own headers and parameters, and the response feeds later steps. Relevance also issues its own API key and ships a JavaScript SDK, so agents can be triggered and read from an application you already run. |
What is missing | We found no native Clay connection, and Relevance lists Clay under Competitors in its own site footer, so joining the two means a webhook or an HTTP step on both sides. There is also no native cold email infrastructure, so Instantly, Smartlead and Lemlist reach it through Pipedream or a custom API step. |
Pro’s
A RevOps person with no engineering help can ship a working agent in an afternoon, because a Tool is assembled from visual steps and an Agent is a prompt plus a list of those tools.
Vendor Credits pass model cost through at wholesale with no markup, and every paid plan lets you plug in your own OpenAI or Anthropic key to bypass that meter completely.
Each run opens up to the individual tool call, with the cost of every tool and the agent's own model spend broken out on the run screen, so a bill that jumps has a traceable cause.
Agents start from more than a thousand app events, scheduled runs, webhooks or direct HTTP requests, so the work fires on the signal rather than waiting for somebody to press go.
The hosted MCP server and the Claude Code plugin let you build and manage agents from Claude Desktop, Cursor or a terminal, and few no-code platforms open that door to engineers.
Phone agents and meeting agents ship with the Team plan instead of a second vendor, so call handling and note taking sit in the same account as the rest of your agents.
Cons
Actions meter every tool run and a failed run still counts against you, so an agent making eight tool calls per account clears the Pro plan's 2,500 monthly actions in roughly 300 accounts.
Top-up actions cost $80 per 1,000, which is around ten times the in-plan rate on Pro at $19 a month for 2,500, so running past your allowance is the expensive way to grow.
The public pricing page now shows a single Enterprise tier behind a sales call, and Free, Pro and Team survive only in the documentation, so the self-serve route is no longer one the site wants you to find.
Concurrency limits are published as Less, Standard and More rather than numbers, and tasks queue once you hit the ceiling, so planning for volume means asking sales instead of reading a spec.
An agent that reasons about which tool to call needs watching before you trust it, and reviewers report docs that shift under them because the product changes so often, so the first month goes on supervision.
Pricing
Free
$0/month No credit card, and free users cannot buy top-ups
200 Actions a month
1,000 Vendor Credits once at signup
Unlimited agents and tools, 1 Workforce
1 build user and 1 project
30-day task history and Marketplace access
Pro
$19/month Billed annually, $29 month to month, for a solo operator
2,500 Actions a month and $20 of Vendor Credits
2 build users and unlimited Workforces
Scheduled tasks, escalations and Chat mode
Premium triggers for WhatsApp, LinkedIn and Telegram
Bring your own LLM keys and A/B testing
90-day task history
Team
$234/month Billed annually, $349 month to month, org-wide not per seat
7,000 Actions a month and $70 of Vendor Credits
5 build users, 45 end users, 5 shared projects
Calling and meeting agents
Analytics dashboard and priority support
Unused credit rollover
Enterprise
Custom The only tier still listed on the public pricing page
Custom Actions and Vendor Credits
Unlimited users and projects
Enterprise triggers for Salesforce, Snowflake and Zendesk
Agent evaluations and work hour controls
SSO, RBAC, audit logs and multi-org management
Dedicated account manager and custom implementation
How it fits your stack
Alternatives
At a glance
No-code agent, tool and workflow builder
Multi-agent Workforces on a visual canvas
About 2,000 app integrations, native plus Pipedream
1,000+ app triggers, webhooks, schedules and API
Phone and meeting agents from the Team plan
Bring your own LLM keys on paid plans
MCP server for Claude, Cursor and ChatGPT
Free plan with 200 actions a month
Read some of our client stories
GTM Workflows
Learn how to use GTM tools in workflows
Guides and tutorials to get the most from your workflows
Ready to get started?
By clicking Sign Up you're confirming that you agree with our Terms and Conditions.
