Melaya vs Dust
Dust builds compliance-grade AI assistants over your company knowledge; Melaya builds approval-gated agents that act, including on real Android apps, with any cloud or local model.
Pick Dust if you want managed, compliance-certified AI assistants over your company knowledge, with mature synced connectors and org-wide governance. Pick Melaya if you need agents that act rather than just answer: operating real Android apps with an on-device approval gate, routing each step across 23 cloud or local model providers, and proving quality with deterministic checks and run traces. Some teams will reasonably run both, Dust for internal knowledge and Melaya for gated real-world actions.
Melaya vs Dust, feature by feature
| Capability | Melaya | Dust |
|---|---|---|
| Mobile app control | Device Control on Android: agents open allowed apps, read the screen, tap, type, and swipe, with an on-device approval gate and app playbooks. No API needed for the app being operated. | None. Agents work in the browser, in Slack, and via API and MCP; no mobile device automation as of 2026. |
| Human-in-the-loop | Approval gates on every consequential write, on the canvas and on the phone before anything is published. | Conversation-first review: people read and use agent output in chat; per-action approval gates are not a headline feature as of 2026. |
| Model choice and local execution | Per-step routing across 23 AI providers, cloud and local via a local runner; bring your own model or keys. | Choice of leading cloud models (Claude, GPT, Gemini, Mistral); inference runs in Dust's hosted cloud. |
| Deployment | Runs in the cloud or fully local on your own hardware. | Hosted SaaS with EU data residency; core platform is MIT open source, but the managed product is cloud-only. |
| Evaluation and observability | Deterministic rule checks before any LLM judge, full run traces, and cost per accepted result. | Usage analytics and agent feedback; no deterministic-first evaluation layer advertised. |
| Connector breadth and company knowledge | ~1,697 tools plus RAG and memory; the catalog skews toward taking actions. | 100+ continuously synced connectors (Slack, Notion, Google Drive, Confluence, Microsoft 365, Salesforce, Zendesk and more) feeding a maintained company knowledge layer. Stronger for internal search. |
| Builder experience | Visual canvas composing agents, tools, triggers, and approval gates; 111 agent personas to start from. | Instruction-based agent builder with attached data sources and tools; quick for non-technical teams. |
| Security and compliance | Open beta; local execution and approval gates reduce data exposure, but no SOC 2 or HIPAA certifications yet. | SOC 2 Type II, GDPR with EU data residency, HIPAA-ready options, SSO (SAML, OIDC) and SCIM, audit logs. |
| Triggers and scheduling | Triggers composed on the canvas; the Forge plan adds scheduled, managed cloud runs. | Agents run on demand in chat and Slack, and programmatically via API and MCP. |
| Pricing | Sandbox free, Outpost $20/mo with your own cloud keys, Forge for scheduled managed cloud. Currently in free open beta. | Credit-metered seats as of mid-2026: Free tier, Pro about $30/seat/mo, Max about $150/seat/mo, custom Enterprise with pooled credits. |
| Best for | Builders and teams who need agents that take gated actions, run on private or local models, or operate mobile apps. | Companies rolling out compliant AI assistants over internal knowledge to many employees. |
Where Melaya wins
- Device Control on Android: agents operate real apps on a real phone, reading the screen and tapping, typing, and swiping, guided by app playbooks, with an on-device approval gate before anything is published. No API is required for the app being operated, and Dust offers nothing comparable.
- Local and private execution: per-step model routing across 23 providers, including local models through a local runner, so sensitive steps never have to leave your hardware. Dust's hosted product runs all inference in its cloud.
- Human-in-the-loop by design: every consequential write passes an approval gate, on the canvas and on the device, which makes action-taking agents safe enough to actually deploy.
- Deterministic-first evaluation: rule checks run before any LLM judge, with full run traces and cost per accepted result, so you can prove an agent works instead of eyeballing outputs.
- Simpler pricing for small teams: a free Sandbox and a flat $20/mo Outpost plan with your own keys, versus credit-metered per-seat pricing.
Where Dust wins
- Enterprise compliance and governance: SOC 2 Type II, GDPR with EU data residency, HIPAA-ready deployment, SSO and SCIM, and audit log retention. Melaya is in open beta and does not carry these certifications yet.
- Company knowledge depth: 100+ continuously synced connectors build a maintained, searchable knowledge layer across Slack, Notion, Google Drive, Confluence, Microsoft 365, Salesforce, and more. For internal Q&A over company data, Dust is the stronger product.
- Org-wide rollout: seat-based administration, Slack-native assistants, and a workspace model designed to put agents in front of every employee, not just builders.
- Vendor maturity: an established enterprise vendor with an MIT open-source core and a security posture that procurement teams recognize.
Melaya vs Dust: frequently asked questions
Is Melaya a good Dust alternative?
Yes, if your goal is agents that take real actions under human approval, run on cloud or local models, or operate Android apps directly. It is not a like-for-like replacement for Dust's company knowledge layer: Dust remains stronger for compliance-certified, org-wide assistants over 100+ synced data sources. Choose based on whether you need agents that act or assistants that answer.
How much does Melaya cost compared to Dust?
Melaya has a free Sandbox tier, a $20/mo Outpost plan where you bring your own cloud keys, and a Forge plan for scheduled managed cloud runs; the platform is currently in free open beta. As of mid-2026 Dust uses credit-metered seats: a Free tier, Pro at about $30 per seat per month, Max at about $150 per seat per month, and custom Enterprise pricing. For a small team paying per seat, Melaya is usually cheaper; large orgs should quote Dust Enterprise directly.
Can Melaya do what Dust does?
Partially. Melaya has RAG, memory, roughly 1,697 tools, and 111 agent personas, so you can build assistants grounded in your own data. It does not match Dust's continuously synced connector fleet or its compliance certifications for org-wide knowledge search. Melaya's strengths point the other way: gated actions, local models, and mobile Device Control.
Can Dust control mobile apps the way Melaya does?
No. As of 2026 Dust agents work in the browser, in Slack, and programmatically via API and MCP; there is no mobile device automation. Melaya's Device Control lets agents operate real Android apps, reading the screen and tapping, typing, and swiping, with an on-device approval gate before anything is published.
Can I run Melaya with local models or fully on-prem?
Yes. Melaya routes each step across 23 AI providers, including local inference through a local runner, and the platform can run fully local. Dust's hosted product is cloud-only, though its core platform is MIT open source for teams willing to operate it themselves.
Which is better for GDPR and enterprise compliance?
Dust, today. It offers SOC 2 Type II, GDPR compliance with EU data residency, HIPAA-ready options, SSO and SCIM, and audit logs. Melaya reduces exposure differently, through local execution and human approval gates, but it is in open beta and does not yet hold those certifications, so regulated procurement will favor Dust.
Do I need API access to the apps my Melaya agents automate?
Not on Android. Device Control operates the real app through the phone's screen: it opens allowed apps, reads what is displayed, and taps, types, and swipes, guided by app playbooks, with an on-device approval gate before publishing. API-based tools are still available whenever an integration exists.
Can I use Melaya and Dust together?
Yes, and it is a reasonable split. Run Dust for compliant internal knowledge assistants across the whole company, and Melaya for approval-gated automations, local-model workloads, and mobile app operations. Both expose programmatic surfaces you can bridge, and neither locks you out of the other.
See it for yourself
Build an agent in the free tier and compare the experience yourself. No card, no waitlist.
Start free