BestOnline.ToolsBestOnline.Tools
Automation and agents

Relevance AI

Relevance AI review: business agents, Workforces and pricing

Relevance AI builds agents and multi-agent Workforces for business tasks. Explore its tools, knowledge, approvals, usage pricing and best alternatives.

Multi-agent teamsCustom toolsHuman approval

Paid plansAccount required

Product previewView original
Relevance AI’s official platform example shows agent, tool, Workforce and knowledge shortcuts alongside recent assistants.
Relevance AI’s official platform example shows agent, tool, Workforce and knowledge shortcuts alongside recent assistants.Relevance AI
Access
Paid access
Example plan
USD 29 / monthPro — monthly
Available on
Web, API
OVERVIEW

Specialist agents connected around business work

Relevance AI is a hosted platform for building agents around business tasks. Each agent combines instructions, tools and knowledge, with a built-in chat interface for requests. Invent can generate an agent from a description, and Marketplace builds provide starting points to clone and adapt. It suits teams turning recurring research, sales or operations work into agents that can use their existing services, rather than simply generate a reply.

Tools give an agent specific actions, such as searching information, enriching a record or sending a message. The visual builder connects inputs, sequential steps and outputs; steps can include model prompts, integrations, API calls and optional Python code. A tool can be reused across agents, shared as a standalone form or run across a knowledge table. Logs show what individual steps received and returned, which helps locate a failing connection or transformation.

Workforces coordinate several specialist agents on a visual canvas. An AI connection lets an agent choose when to hand work to another specialist, while a Next step connection always starts the next agent when its predecessor finishes. Conditional routes and tool chains add more control. The same specialist can participate in multiple Workforces, making a shared research or enrichment agent useful across several processes.

Knowledge gives agents access to business documents and data. Files and connected sources such as Google Drive, SharePoint and Notion can supply material. A small knowledge set can be included directly in the prompt, or an agent can search it with semantic retrieval for relevant passages. Tools can enrich knowledge tables with summaries, tags or extracted fields. The usefulness of those answers still depends on the source material and how retrieval is configured.

Native integrations supply actions and triggers for supported apps. A Pipedream connection extends access to additional services, and custom API keys support endpoints beyond the built-in catalogue. These routes differ: a broad integration listing does not mean every service has the same ready-made steps or triggers. Plan access also matters, with Salesforce, Snowflake and Zendesk triggers listed under Enterprise.

Recurring schedules run tasks daily, weekly, monthly or on a custom cadence, using the account’s configured timezone. Scheduled messages are separate: an agent can arrange a future follow-up or reminder when that feature is enabled. The settings can cap daily scheduled actions and delay messages while a task is active. App events, webhooks, the interface and API calls provide other ways to start work.

Approval Mode is configured separately for each tool. Autopilot runs it directly; Human-in-the-Loop waits for explicit approval, and an agent can mix both modes. Task and tool histories help review activity and failures. Organization-level usage views separate Action and Vendor Credit consumption from concurrency, which measures simultaneous tasks. When concurrent capacity is reached, further tasks queue rather than representing another type of usage credit.

The API and Python or JavaScript SDK let application code trigger agents, run tools and manage resources. MCP works in two directions: an agent can use remote servers supporting Streamable HTTP, while Relevance AI’s own server exposes project capabilities to compatible external clients. Local MCP servers on a user’s machine are not supported by the agent client. Project roles and Run-Only mode limit editing through the Relevance server; they do not turn all existing agent actions into read-only lookups. The platform is hosted, and connected external providers remain part of processing.

Best for

  • Sales and operations teams building specialist business agents
  • Teams coordinating research, enrichment and multi-step handoffs
  • Builders combining reusable tools, business knowledge and human approval

Limitations

The Free plan has been retired. Failed tool runs still consume Actions, and model or external-tool costs are separate. Pro and Team limit build users and projects; calling and meeting agents require Team or Enterprise, and some triggers and evaluations are Enterprise-only. Concurrent tasks above plan capacity queue. Human approval requires configuration, and connected services have their own data-processing scope.

WHAT YOU CAN DO

Relevance AI features

Business agents and reusable starting points

Build an agent from instructions, use Invent from a description, or adapt a Marketplace agent to your team’s work.

Visual multi-agent Workforces

Combine specialist agents through AI handoffs, fixed next steps and conditional paths, with reusable agents across processes.

Custom tools and optional code

Chain model prompts, app actions and API calls into reusable tools, with Python available for custom logic.

Business knowledge and semantic retrieval

Connect documents and data, include small sets in the prompt or search larger collections, and enrich tables with tools.

App events and scheduled work

Start tasks from integrations, webhooks or schedules. Scheduled messages add future reminders and follow-ups when enabled.

Per-tool human approval

Mix tools that run automatically with tools that wait for explicit approval inside the same agent.

Run logs, usage and capacity views

Review tool steps and agent costs, track the two usage pools, and see concurrent tasks against plan capacity.

API, SDK and two MCP directions

Call Relevance from code or external clients, and connect agents to compatible remote MCP servers with project-scoped access.

PLANS & PRICING

Relevance AI pricing

Relevance AI has paid Pro and Team organization plans, plus custom Enterprise pricing. Pro starts at $29 month to month or $19 per month equivalent with annual billing. Subscriptions include Actions and Vendor Credits: an Action is a tool run, while Vendor Credits cover separate model and tool costs. Bringing your own provider keys bypasses Vendor Credits but leaves Action billing and provider charges. Included Actions reset at renewal; purchased Actions carry forward, and Vendor Credits roll over while subscribed. Additional usage, build users, projects and plan-specific features affect the total cost.

Example plans.

Pro — monthly

USD 29 / month

Relevance AI organization

Plan details

Organization plan; 2 build users/1 project; 2,500 Actions and USD 20 Vendor Credits/month.

Pro — annual commitment

USD 19 / month

Relevance AI organization

Plan details

Monthly equivalent billed annually; 30,000 Actions and USD 240 Vendor Credits/year; 2 build users/1 project.

Team — monthly

USD 349 / month

Relevance AI organization

Plan details

Organization plan; 5 build users/45 end users/5 projects; 7,000 Actions and USD 70 Vendor Credits/month.

Pricing updated 2026-10-10. Prices may change.View current pricing →
Technical specifications

Automation and agents

FeatureRelevance AI
Connectors and tool access
Native app actions/triggers, Pipedream connections, custom API steps and remote Streamable HTTP MCP tools. Reusable tools combine prompts, integrations and optional code.Credentials, native vs Pipedream support and plan-specific trigger access
Triggers and scheduling
UI/API requests, app events, webhooks and recurring schedules. Scheduled messages must be enabled; schedule timezone and daily limits are configurable.Configured triggers; Salesforce/Snowflake/Zendesk triggers are Enterprise-specific
Human approval controls
Autopilot permits immediate tool use; Human-in-the-Loop waits for explicit approval. Different tools in an agent can use different modes.Selected tools must be configured; no universal approval guarantee
Hosting and deployment
Hosted platform with authenticated access, API/SDK and project-scoped MCP connections. Connected third-party processing falls outside the platform security policy.Hosted service; Enterprise adds organization controls; provider terms remain relevant
Execution accounting
Organization subscription plus Actions per Tool run, including failures, and separate Vendor Credits for model/tool costs. Own keys bypass Vendor Credits, not Actions or provider charges.Included Actions reset; purchased Actions and subscribed Vendor Credits carry forward
API access
SupportedPython/JavaScript SDK and project-scoped platform API; external service keys have a separate purpose
MORE AI TOOLS

Relevance AI alternatives

Explore platforms for connected agents, AI applications and broader business automation.

Gumloop

Gumloop

Connected team agents with reusable skills and model/tool/compute credit billing.

Dify

Dify

Visual AI applications and knowledge workflows with Cloud or Community deployment.

n8n

n8n

Connected automation and AI-agent workflows with Cloud or self-hosted operation.

COMMON QUESTIONS

Relevance AI FAQs

Does Relevance AI have a free plan?

The Free plan is retired for new signups and existing accounts. The current documented plans are Pro, Team and custom Enterprise; an organization previously on Free needs a paid plan to continue.

How much does Relevance AI cost?

Pro is $29 per organization month to month, or $19 per month equivalent billed annually. Team is $349 monthly or $234 per month equivalent billed annually. Included Actions and Vendor Credits are separate allowances, with additional usage available. Enterprise pricing is custom.

What are Actions and Vendor Credits?

An Action is one tool run, including a failed run. Vendor Credits cover the separate model and tool usage costs. Your own provider keys can bypass Vendor Credits, but Actions still apply and the provider can charge for usage. Included Actions reset at renewal; purchased Actions carry forward and Vendor Credits roll over while subscribed.

What are Relevance AI Workforces?

Workforces connect specialist agents and tools on a visual canvas. AI connections allow an agent to choose a handoff, Next step connections enforce a sequence, and conditional paths route work according to configured rules. An agent can be reused in more than one Workforce.

Can Relevance AI use company documents?

Yes. Knowledge sources can include uploaded files and connected services such as Google Drive, SharePoint and Notion. An agent can receive a small knowledge set directly in its prompt or search a larger collection through semantic retrieval. The documents and retrieval setup affect the quality of the answers.

Can Relevance AI require human approval?

Yes. Human-in-the-Loop mode requires explicit approval before a selected tool runs. Other tools can remain in Autopilot, so approval is a per-tool setting rather than a guarantee that every action is reviewed.

Does Relevance AI support MCP?

Yes. Its agent client connects to remote Streamable HTTP servers, not local servers on your machine. Its own MCP server lets compatible external clients access a Relevance project. Project roles and optional Run-Only mode restrict project-editing capabilities; ordinary agent and tool usage still follows the plan’s billing.

What are the best Relevance AI alternatives?

Gumloop combines connected agents and team automation. Dify focuses on AI applications and knowledge workflows with Cloud or Community deployment. n8n combines app automation and AI with Cloud or self-hosted operation. Relevance AI stands out for coordinating specialist business agents through Workforces and reusable tools.

Is this your product? Claim this page to update your listing.