Comparison
ResolveKit vs Intercom Fin
Both promise AI-powered support. The difference is architectural: Intercom Fin operates as a web widget outside your product, while ResolveKit embeds natively inside your app — where users actually get stuck.
| Feature | Intercom Fin | ResolveKit |
|---|---|---|
| Architecture | Web widget / help center overlay | Native embedded SDK (iOS + Android) |
| Where the agent lives | Outside your product (browser overlay) | Inside your app — native screen, same context |
| Product context | Limited — sees only what the widget can access | Deep — SDK has access to app state, screen, version, workflows |
| Can take action | No — answers questions only | Yes — executes actions with approval guardrails |
| Approval system | N/A — no actions to approve | Policy-based approvals, user consent, operator control |
| Session traces | Conversation logs only | Full traces: context, proposals, approvals, outcomes |
| Open source | No — proprietary SaaS | Yes — MIT SDKs, AGPL backend, fully auditable |
| Self-host option | No | Yes — run on your own infrastructure, bring your own LLM keys |
| Mobile support | Mobile web widget only | Native iOS (SwiftUI + UIKit) + Android (Compose + Views) |
| LLM flexibility | Fixed — Intercom's Fin AI | Any LLM via LiteLLM — OpenAI, Claude, Gemini, local models |
| Data residency | Intercom-controlled regions | Your infrastructure — full data control |
| Compliance | SOC 2, GDPR (vendor-managed) | You control compliance via self-hosting |
| Analytics & insights | Intercom dashboard only | Full session traces exportable for custom analytics |
| Pricing model | ~$0.99 per resolution | ~$0.05/resolution typical (Gemini Flash-Lite, BYO) + $0 platform fee |
| Cost (self-hosted) | Not available | Free — open source, bring your own LLM keys |
| Seat fees | Yes — per-agent pricing on top of resolution costs | No — no platform seat fees |
Key differences
Where the agent lives
Intercom Fin operates as a web widget or inbox assistant. It sees what the user types and what is on the page. ResolveKit embeds directly in your app via SDK — it understands the user's exact screen, workflow, and app state. On mobile, this is the difference between a widget covering half the screen and a native chat view that feels like it belongs in your app.
Can actually take action
Intercom Fin answers questions and surfaces knowledge base articles. ResolveKit can explain, guide, and — with approval — execute actions like resetting a password, applying a refund, or navigating the user to the right screen. Every action goes through policy-defined approval checkpoints, giving operators real control over what the agent can do.
Cost at scale
Intercom Fin costs approximately $0.99 per resolution, plus per-agent seat fees. ResolveKit has no platform fee per resolution or per seat. With bring-your-own keys, teams often land around $0.05 per resolution on Gemini Flash-Lite. You can self-host or use ResolveKit-managed infrastructure while keeping model spend on your own keys.
Open source and self-hosting
Intercom is a proprietary SaaS platform — you cannot inspect, audit, or modify the system interacting with your users. ResolveKit is fully open source: MIT license for SDKs, AGPL for the backend. You can self-host the entire stack on your own infrastructure with any OpenAI-compatible LLM.
Can you use both?
Yes — they solve different parts of the funnel. ResolveKit catches issues before they become tickets by resolving them in-app. Intercom excels at managing the tickets that do get created. Teams often use ResolveKit as the first line of defense and Intercom as the escalation path for complex cases that need human agents.
Which should you choose?
Choose Intercom Fin if...
- • Your primary need is web-based support with an existing help desk
- • You already use Intercom for marketing, sales, and support
- • Your product is web-only with no native mobile app
- • You don't need AI agents to execute actions in your product
Choose ResolveKit if...
- • You have a native iOS or Android app with significant support volume
- • Users get stuck in-product and leave before contacting support
- • You want AI that can actually resolve issues, not just explain them
- • You need approval guardrails and full audit trails for AI actions
- • Open source and self-hosting are requirements, not nice-to-haves