Pricing Landscape in Agent Runtimes
On this page
Research date: June 2026 Scope: Competitive pricing analysis of 6 agent runtime / AI workflow platforms, focused on pricing parameters, overage mechanics, and execution definitions.
1. Executive Summary
The agent runtime market prices on three primary dimensions: executions (workflow/agent runs), seats (builders and end-users), and features (tier-gated capabilities). Most platforms offer free tiers with low usage caps, mid-tier plans for teams ($20-350/mo), and enterprise custom pricing. The definition of an “execution” varies significantly across competitors — ranging from a single tool call (Relevance AI “Actions”) to an entire end-to-end workflow run (n8n “executions”). This divergence makes direct cost-per-unit comparison unreliable; the pricing philosophy must be understood per-platform.
Overages are handled inconsistently: some platforms allow ad-hoc top-ups (Relevance AI), some offer per-unit overage pricing (n8n, Pinata), and others pause service until upgrade or plan renewal (Zapier). Most self-serve plans allow buying more usage without upgrading tiers.
2. Competitor Pricing Profiles
2.1 Relevance AI
Source: https://relevanceai.com/pricing (confirmed June 2026)
| Plan | Monthly Price | Executions (Actions) | Seats | Key Features |
|---|---|---|---|---|
| Free | $0 | 200 Actions + $2 credits | 1 build user, 1 project | Unlimited agents & tools, SOC 2 Type II |
| Team | $349 | 7,000 Actions + $70 credits | 5 build users, 45 end users | Shared projects, analytics, calling/meeting agents |
| Enterprise | Custom | Custom | Custom | SSO, RBAC, audit logs, dedicated support |
- What is an execution? An Action is a single tool invocation by an agent (e.g., one API call, one LLM prompt, one data lookup). Vendor Credits separately cover the underlying LLM model costs (OpenAI, Anthropic, etc.) at no markup. On paid plans, you can BYO API keys to bypass Vendor Credits entirely.
- Overage/refill? Yes. Top-up packs available for both Actions and Vendor Credits without upgrading the plan. This makes costs flexible but less predictable at scale.
- Notable: Unlimited agents on all plans (cost is purely usage-driven, not agent-count-driven). SOC 2 Type II, GDPR.
2.2 Pinata Cloud
Source: https://pinata.cloud/pricing (June 2026)
| Plan | Monthly Price | Agents | Storage | Gateways |
|---|---|---|---|---|
| Free | $0 | 0 | 1GB | 1 |
| Picnic | $20 | 1 | 1TB | 1 + CDN |
| Fiesta | $100 | 3 | 5TB | 3 + CDN |
| Enterprise | Custom | Custom | Custom | Custom |
- What is an execution? Pinata prices on storage, bandwidth, and requests — not per-agent-execution. An “agent” is a billable entity (distinct autonomous file-storage agent). Requests are API calls to storage/gateway infrastructure.
- Overage/refill? Yes. Extra storage ($0.07/GB Picnic, $0.035/GB Fiesta), bandwidth ($0.10/GB, $0.08/GB), and requests ($0.20/10k, $0.15/10k) all available as add-ons without plan upgrade.
- Notable: Pinata is fundamentally an IPFS/file-storage platform with agents as a secondary offering; pricing is storage-first, not execution-first. Less relevant as a direct agent runtime competitor — included for completeness as it prices per “agent” differently.
2.3 Gumloop
Source: https://www.gumloop.com/pricing (June 2026)
| Plan | Monthly Price | Credits | Seats | Concurrent Runs | Agent Interactions |
|---|---|---|---|---|---|
| Free | $0 | 5,000/mo | 1 | 2 | 5 |
| Pro | $37 | 20,000+/mo | Unlimited | 5 | 25 |
| Enterprise | Custom | Custom | Custom | Custom | Custom |
- What is an execution? Gumloop uses credits as its pricing unit. The exact definition of a credit is not publicly documented in detail on the pricing page (FAQ question “What is a credit?” exists but content not expanded in fetched page). Based on platform mechanics, credits appear to be consumed per-agent-interaction and tool use. A separate “Agent Interactions” limit governs concurrent agent conversations. BYO API keys supported.
- Overage/refill? The plan comparison table lists “Credit Overages” as a feature gated behind Enterprise (not available on Free/Pro). This implies Pro users hitting the credit cap must upgrade to Enterprise for overages, rather than buying top-ups. Annual billing offers 20% discount.
- Notable: Unlimited agents and unlimited workflows on all paid plans. Credit slider on pricing page goes up to 1.5M credits for higher-volume teams. “Concurrent Runs” limits how many workflows execute simultaneously. Enterprise includes RBAC, SCIM/SAML, audit logs, VPC.
2.4 n8n
Source: https://n8n.io/pricing/ (June 2026)
| Plan | Monthly Price (annual) | Executions/mo | Users | Concurrent | Notable |
|---|---|---|---|---|---|
| Starter | €20 | 2,500 | Unlimited | 5 | Cloud-hosted, 1 project |
| Pro | €50 | 10,000 | Unlimited | 20 | 3 projects, 7d insights |
| Business | €667 | 40,000 | Unlimited | — | Self-hosted, SSO/SAML/LDAP, Git |
| Enterprise | Custom | Custom | Unlimited | 200+ | On-prem/cloud, SLA, external secrets |
- What is an execution? An execution is a single full run of an entire workflow from start to finish — regardless of how many steps/nodes are inside it. n8n explicitly markets this as its pricing advantage: “Unlike other tools that charge per step or user, n8n lets you build freely and only pay when a workflow runs.” A single n8n execution could encompass work that would be 10,000 “operations” in other tools (per their marketing).
- Overage/refill? Yes. Business plan overage: €4,000 for 300,000 extra executions. Enterprise has custom overage. Workflows continue running after quota exhaustion; sales team contacts you when approaching 80% of annual quota. Overage invoiced 45 days after occurrence. n8n also offers a free, self-hosted Community Edition (open source) — effectively unlimited executions if you host it yourself, with no paid features.
- Notable: Unlimited users on all plans. Self-hosting option changes the value equation dramatically. AI Workflow Builder credits are separate from execution count (50/150/1000 credits per plan for AI-assisted building). Enterprise plan available either cloud-hosted or self-hosted.
2.5 Zapier
Source: https://www.lindy.ai/blog/zapier-pricing (via Lindy’s analysis, May 2026 — Zapier’s own pricing page unreachable)
| Plan | Monthly Price | Tasks/mo | Users | Key Features |
|---|---|---|---|---|
| Free | $0 | 100 | 1 | Two-step Zaps only |
| Professional | $29.99 | 750 | 1 | Multi-step Zaps, filters, 8,000+ apps |
| Team | $103.50 | 2,000 | Unlimited | Shared folders, admin, 2-min update time |
| Enterprise | Custom | Custom | Unlimited | SSO, SCIM, audit logs, dedicated CSM |
- What is an execution? A task is each individual action step within a Zap. A single Zap that triggers on a new email, filters it, and creates a CRM record = 2+ tasks. Tasks are granular (per-step), making Zapier one of the most expensive at scale per unit of automation.
- Overage/refill? Limited. Zaps pause when task limit is exceeded until next billing period or plan upgrade. Task bundles can be purchased on paid plans. Very high-volume tiers reach $5,999/month for 2M tasks.
- Notable: Largest integration library (8,000+). No AI-agent-native pricing — Zapier’s agent features (Zapier Agents) are separate from core Zap pricing. The pricing model is task-count-based, not agent-based. No native “bring your own API key” model for LLM costs.
2.6 StackAI
Source: https://www.stackai.com/pricing (June 2026)
| Plan | Monthly Price | Runs/mo | Projects | Seats |
|---|---|---|---|---|
| Free | $0 | 500 | 2 | 1 |
| Enterprise | Custom | Custom | Unlimited | Custom |
- What is an execution? A run is a single execution of an agent workflow or flow. Each time an agent processes input through its pipeline (LLM calls, data loaders, logic steps), that consumes one run.
- Overage/refill? Enterprise plan has custom limits, so overage is negotiated. The Free plan has a hard cap of 500 runs; no self-serve top-up path visible.
- Notable: Recently acquired by Asana (announced June 2026). Enterprise-only commercially — no self-serve paid tier between Free and Enterprise. Enterprise includes on-prem deployment, VPC, SSO, SOC 2 Type II, HIPAA, GDPR, ISO 27001, dedicated solution engineers. “Run” is the only pricing unit.
3. Cross-Competitor Comparison
3.1 How Each Platform Defines “an Execution”
| Platform | Unit Name | Definition | Granularity |
|---|---|---|---|
| Relevance AI | Action | Single tool invocation by an agent (LLM call, API call, data operation) | Fine-grained (per-step) |
| Pinata Cloud | Request | API call to storage/gateway infrastructure | Infrastructure-level |
| Gumloop | Credit | Consumption unit for agent interactions/tool use (exact formula undisclosed) | Opaque credit system |
| n8n | Execution | Full workflow run from start to finish, regardless of steps | Coarse-grained (whole workflow) |
| Zapier | Task | Individual action step within a Zap | Fine-grained (per-step) |
| StackAI | Run | Single agent workflow execution through its full pipeline | Medium-grained (per-pipeline) |
Key insight: The market has no standard unit. Relevance AI and Zapier charge per “step” (expensive at scale), n8n charges per “workflow” (cheaper for complex automations), and Gumloop uses opaque “credits.” This makes direct cost comparison difficult and is a competitive positioning lever.
3.2 Pricing Parameters Summary
| Platform | Free Tier Limit | Entry Paid | Mid-Tier | Enterprise | Per-Seat Pricing? |
|---|---|---|---|---|---|
| Relevance AI | 200 Actions | — | $349/mo (Team) | Custom | Yes (5 build + 45 end on Team) |
| Pinata Cloud | 0 agents | $20/mo (Picnic) | $100/mo (Fiesta) | Custom | No (per-agent, not per-seat) |
| Gumloop | 5,000 credits | $37/mo (Pro) | — | Custom | No (unlimited seats on Pro) |
| n8n | Community (free) | €20/mo (Starter) | €50/mo (Pro) | €667-€Custom | No (unlimited users all plans) |
| Zapier | 100 tasks | $29.99/mo (Pro) | $103.50/mo (Team) | Custom | No (per-workspace, not per-seat) |
| StackAI | 500 runs | — | — | Custom | Yes (custom seats on Enterprise) |
3.3 Overage/Refill Mechanics
| Platform | Can buy more without upgrading? | Mechanism |
|---|---|---|
| Relevance AI | Yes | Top-up packs for Actions + Vendor Credits |
| Pinata Cloud | Yes | Add-on storage, bandwidth, requests at per-unit prices |
| Gumloop | Enterprise only | Credit overages gated behind Enterprise plan |
| n8n | Yes | Overage invoiced (€4,000/300K execs on Business); self-hosted Community is free |
| Zapier | Yes (on paid plans) | Task bundles; Zaps pause otherwise |
| StackAI | Unknown (Enterprise custom) | Custom via enterprise agreement |
4. Implications for FourIA Pricing
Note (July 2026): This analysis was written based on an earlier pricing model (seats + feature tiers + credits). FourIA v1.0-beta has since adopted a flat-fee credit-based plan (175€+VAT/month, 9,000 credits, 5 builder seats, unlimited reader users). Credit ≠ AI expense. The competitive insights below on credit models, overage mechanics, and BYO key patterns remain relevant.
- Gumloop’s unlimited-seats model on paid plans + credit system (though Gumloop gates overages behind Enterprise)
- Relevance AI’s Actions + Vendor Credits split (though Relevance AI charges per-action granularly)
- n8n’s execution-based model where “execution = full workflow” (the most cost-predictable model for users)
Recommendations from this research:
- Define your “execution” unit early and defend it. n8n’s “full workflow = 1 execution” is the most user-friendly and predictable model. Relevance AI’s per-Action model creates cost anxiety. Gumloop’s opaque “credit” creates confusion.
- Free tier should cover testing. All competitors offer free tiers (100-5,000 units). 5,000 is the high watermark (Gumloop). 100-500 is typical for workflow platforms.
- Overage without tier upgrade is a differentiator. Relevance AI does this well. n8n’s Community Edition (free self-hosted) is the ultimate “unlimited free tier” — a powerful acquisition funnel.
- Seat-based pricing is the minority pattern. Most platforms charge per-workspace or per-execution, not per-seat. n8n and Gumloop offer unlimited users on paid plans. Relevance AI is the notable exception with distinct build-user and end-user seat limits. FourIA’s seat-based approach is a differentiator if paired with generous execution limits.
- Vendor Credits / BYO API keys is a growing pattern. Relevance AI and Gumloop both support BYO keys to let customers control LLM spend directly. This reduces platform margin risk and gives customers cost control — a compelling feature for FourIA to adopt.
5. Sources
- Relevance AI pricing: https://relevanceai.com/pricing (June 2026)
- Relevance AI plan details: https://www.lindy.ai/blog/relevance-ai-pricing (May 2026)
- Pinata Cloud pricing: https://pinata.cloud/pricing (June 2026)
- Gumloop pricing: https://www.gumloop.com/pricing (June 2026)
- n8n pricing: https://n8n.io/pricing/ (June 2026)
- Zapier pricing: https://www.lindy.ai/blog/zapier-pricing (May 2026 — Zapier’s own page unreachable)
- StackAI pricing: https://www.stackai.com/pricing (June 2026)