Workato
The strongest protocol and guardrail story on this table so far: MCP in both directions, A2A out to four external agent frameworks, PII stripped before the model ever sees it. You still cannot find out what any of it costs without a sales call, and the median buyer pays about €59,000 a year.
PRICING
| Base workspace | Quote only | Platform fee. Independent estimates start around $10,000/yr for the smallest deployments |
| Business edition | Quote only | Most common mid-market tier. Vendr-observed $61,800–$78,500/yr at 5M tasks/yr against a ~$120,000 list |
| Enterprise edition | Quote only | Adds full API management, compliance and 24/7 support. Vendr-observed $84,200–$128,300/yr at 5M tasks/yr |
| Workato ONE | Quote only | Top edition. Everything in Enterprise plus the agentic capabilities: Agent Studio, Enterprise MCP, Workato GO |
| Add-ons | Quote only | Recipe packs, high-volume recipes at a flat rate instead of per task, on-premise agent, custom connectors, extra seats |
Usage is metered in business actions, not seats. Failed actions are excluded; rerun jobs are counted. A parent job does not roll up the usage of the recipes it calls, so each child job bills separately and cost forecasting on nested recipes is harder than it looks. AI by Workato carries its own rate limits: 60 requests a minute for direct customers, 10 for embedded partner workspaces, 5 and a 10,000-token ceiling on self-service workspaces.
checked 2026-08-21 · vendor pricing page
Element scores
Strengths
Protocols and governance are where this platform earns its score. Enterprise MCP runs in both directions with OAuth, per-server rate limits, identity-aware execution and audit logging, hosted in US, EU and APAC regions with data residency honoured, and an A2A connector reaches Google ADK, LangGraph, Bedrock AgentCore and Azure AI Foundry agents. Agent Guardrails block, redact or tokenise PII before the model sees it and route high-stakes actions to a human in Slack or Teams. Underneath that sits the thing Workato has spent a decade building: a connector library deep enough that a skill is just a recipe, plus a Ruby Connector SDK with a local CLI, unit tests and git so custom connectors get real version control. Grounding is properly designed, with knowledge bases for semantic questions and Data tables for exact lookups and a documented rule for routing between them.
Honest dings
You cannot price this. The pricing page published nothing at all on 21/08/2026, the median tracked contract is about €59,000 a year, and usage bills in business actions where nested child jobs meter separately, so the forecast you build before signing will be wrong in a direction you cannot predict. The managed AI connector is dated: the docs still name Anthropic Sonnet 4 as the model in most regions, so BYOLLM is effectively mandatory to get current models. There is no way to measure output quality either, only whether a workflow ran: Test mode gives you scenarios, not scores, so a prompt change ships without a regression gate. And independent reviewers keep hitting the same data ceilings, with CSV parsing capped at 50,000 records and lookup tables at 100,000 rows.
Sources (22) — every claim traceable
Every audit lists the research it rests on — transparency and traceability are the product. Tools evolve: each audit is a snapshot of its audit date, and re-audits supersede older versions (kept below for reference).
- workato.com/pricing — Live pricing page rendered 21/08/2026: no rates, no tiers, no free plan, only a Schedule a demo call to action (accessed 2026-08-21)
- docs.workato.com/en/agentic/agentic.html — Official agentic docs: genies, agent registry and agent cards, skill and environment version management, large action models, multi-modal input types, session and persistent memory, decision models, Assign task to genie, A2A Protocol connector naming Google ADK, LangGraph, Bedrock AgentCore and Azure AI Foundry (accessed 2026-08-21)
- docs.workato.com/en/mcp.html — Enterprise MCP: hosted, local and Developer API MCP servers, dynamic discovery, OAuth authentication, server-level rate limiting, identity-aware execution and audit logging, availability across US/EU/AU/JP/SG/IL/KR/UK data centres and exclusion of the CN data centre (accessed 2026-08-21)
- docs.workato.com/en/agentic/agent-studio — Genie components table (AI model, chat interface, knowledge base, skills), RBAC and verified user access (accessed 2026-08-21)
- docs.workato.com/agentic/agent-studio/knowledge-bas… — Knowledge bases as the genie's memory: semantic-similarity retrieval, metadata filtering by created date, source and knowledge base ID, knowledge recipes syncing from connected apps (accessed 2026-08-21)
- docs.workato.com/agentic/agent-studio/knowledge-bas… — Hybrid semantic search confirmed, plus the documented routing rule separating knowledge bases (semantic) from Data tables (exact lookup and aggregation) (accessed 2026-08-21)
- docs.workato.com/en/agentic/agent-studio/test-genie… — Test mode: isolated live environment, saved scenarios with custom prompts, guidance on happy paths, edge cases and out-of-scope requests. No output scoring surface (accessed 2026-08-21)
- docs.workato.com/en/agentic/agent-studio/connectors… — Assign task to genie action: conversation ID for chained calls, task metadata as datapills, up to ten additional context files, structured output configuration (accessed 2026-08-21)
- docs.workato.com/connectors/ai-by-workato — Managed AI connector: processes data using Anthropic Sonnet 4 in most regions, does not train on user data, US and EMEA data residency, rate limits of 60/10/5 requests per minute by workspace type and a 10,000-token cap on self-service (accessed 2026-08-21)
- workato.com/product-hub/changelog/bring-your-own-ll… — BYOLLM announcement, January 2026: power genies with your own OpenAI or Anthropic credentials instead of Workato's managed contracts (accessed 2026-08-21)
- finance.yahoo.com/technology/ai/articles/workato-in… — Business Wire, 09/07/2026: Headless API and Agent Guardrails. PII blocked, redacted or tokenised before the model; denied topics and blocked words; identity or approved service account; human approval in Slack or Teams; auto-redacted Conversation History; SOC 2 Type II, ISO 27001, HIPAA and PCI-DSS 4.0; model flexibility across OpenAI, Anthropic and AWS Bedrock on customer credentials (accessed 2026-08-21)
- helpnetsecurity.com/2026/07/10/workato-headless-api… — Independent coverage of the same launch: Genies responding to business events without a human trigger, running inside any agent harness, per-genie access scoping and instant revocation by key rotation (accessed 2026-08-21)
- docs.workato.com/en/pricing/pricing-faqs.html — Official billing rules: usage-based, tracked separately per product capability, failed actions excluded, rerun jobs counted, parent jobs do not roll up child job usage (accessed 2026-08-21)
- docs.workato.com/pricing/usage-metrics/recipes.html — Business actions as the billing unit: only triggers and app action steps on eligible apps count; conditionals and monitoring blocks do not (accessed 2026-08-21)
- docs.workato.com/developing-connectors/sdk/cli.html — Connector SDK Ruby gem and CLI: local emulation of the Workato runtime, unit tests, push to git for team review (accessed 2026-08-21)
- finance.yahoo.com/news/workato-named-leader-2026-ga… — Business Wire, 18/03/2026: Leader in the 2026 Gartner Magic Quadrant for iPaaS for the eighth consecutive time, Furthest in Completeness of Vision for the third year, 4.9/5 across 555 Gartner Peer Insights reviews as of 17/03/2026, dateline Palo Alto (accessed 2026-08-21)
- vendr.com/marketplace/workato — Independent procurement data: median $64,544 per year across 340 purchases, 19% average savings, tier ranges by task volume, professional services $10,000–$50,000+, premium connectors often extra (accessed 2026-08-21)
- zapier.com/blog/workato-pricing — Competitor-published but Vendr-sourced pricing table: $61,800–$78,500 Business and $84,200–$128,300 Enterprise at 5M tasks/yr, plus the two-part platform-fee-plus-usage model (accessed 2026-08-21)
- integrate.io/blog/workato-review — Independent review with the concrete data ceilings: CSV parsing capped at 50,000 records, lookup tables at 100,000 rows, recipe-based pricing escalating at scale, G2 reviewers calling pricing opaque and expensive (accessed 2026-08-21)
- activepieces.com/blog/workato-pricing — Edition ladder including Workato ONE as the tier that carries the agentic capabilities, plus negotiated-versus-list comparisons at 5M and 10M tasks (accessed 2026-08-21)
- unifygtm.com/insights-headcount/workato — 35% year-on-year ARR growth for the fiscal year ended 31/01/2026, named customers Samsara, Vodafone, Lucid Motors and Atlassian (accessed 2026-08-21)
- getlatka.com/companies/workato — Company profile: roughly 1,400 employees in 2026, $415M raised across five rounds, $5.7B valuation from the 2021 Series E, CEO Vijay Tella, founded 2013 (accessed 2026-08-21)