Pinned official revision
The deployment tracks a reviewed upstream release rather than a moving latest tag.
Build production AI applications with visual workflows, RAG pipelines, tools, model management, and APIs. Darwa adds the managed runtime, HTTPS, persistent storage, and lifecycle tracking around it.
Dify is the closest thing to a full stack for AI products: prompt orchestration, a RAG pipeline, a dataset manager, tool calling, and a hosted API and web app around whatever you build.
The visible capabilities come from the verified catalog entry; secrets and signed installer assets remain private to the backend.
The deployment tracks a reviewed upstream release rather than a moving latest tag.
Each service in the Compose stack receives its own generated credential.
Every container in the stack is recorded in the backend, not just the entrypoint.
A working address with a valid certificate from the first health check, no manual TLS setup.
Chunking, embedding, retrieval, and grounded answers are managed as one pipeline.
Register providers, models, and tools once and reuse them across every application.
A selection of what the upstream project supports. Credentials for any third-party service are yours to add and revoke.
Dify ships as a Compose stack. Darwa pins the revision, generates every secret, and tracks each service as part of one deployment.
Create conversational and workflow applications with a managed API and web interface.
Build retrieval pipelines over approved documents with application-level orchestration.
Combine models, tools, datasets, and team workflows inside one controlled environment.
The hosting layer is the same for every agent, which is what makes a one-click install honest.
Deployments run from a reviewed catalog entry rather than an arbitrary image you have to trust.
Each agent gets its own workload boundary, not a shared container with an unrelated website.
Plan storage is allocated up front so application state survives restarts and releases.
A working address with a certificate from the first health check, with custom domains where supported.
Runtime and resource signals come from the actual workload, so a failure is visible rather than silent.
Grace, suspension, retention, and restoration follow the plan instead of an untracked server.
One AI Agent subscription covers the managed runtime, storage, and endpoint. Model provider usage is billed by that provider.
Live pricing is temporarily unavailable. Dify needs at least 1 vCPU, 4 GB RAM, and 20 GB of storage.
View AI agent plansDarwa keeps Dify selected when the authenticated launch flow opens.
Use an active agent subscription or choose a plan with enough compute and storage.
Start with an agents.darwa.co address or select an eligible verified custom domain.
The backend verifies the manifest, creates the runtime, and records its service and container identities.
Darwa hosts the open-source application independently; upstream licensing and product behavior remain with its project.
Dify is oriented toward shipping a finished application: it bundles a dataset manager, a hosted API, a web app, and usage logging. Langflow is oriented toward designing and testing the graph itself. Teams often prototype in Langflow and productionise in Dify.
It runs several services at once, including a worker and a vector store, and indexing a dataset is memory-intensive. The 4 GB catalog minimum reflects the whole stack rather than a single container.
Yes. Each application gets an API endpoint and an optional hosted web interface, both reachable over the managed HTTPS address, with API keys you issue and revoke from the console.
Uploaded files and their embeddings are written to the deployment's persistent storage, so they survive restarts and releases within the plan's allocated capacity.
Darwa manages the verified deployment manifest, isolated runtime, persistent storage allocation, HTTPS endpoint, health tracking, usage visibility, and paid-plan lifecycle. You manage the agent's models, credentials, tools, content, and decisions.
Choose Install Dify, sign in, and select an AI Agent plan if the workspace does not already have one. The launch flow keeps Dify selected while you choose its address and start the managed deployment.
The current catalog minimum is 1 vCPU, 4 GB RAM, and 20 GB of storage. Real requirements depend on concurrency, models, document volume, integrations, and enabled features.
Darwa allocates persistent storage according to the selected plan. Important application data should still follow the backup and recovery practices required by your business.
The application can start with a managed Darwa address. A domain you own can be selected when custom-domain support is available for the launch flow and its DNS is verified.
No affiliation is implied. Dify is a third-party open-source project. Darwa provides independent managed hosting and links to the upstream repository for project information and licensing.
Sign in, pick a plan with enough resources, and Darwa builds the managed runtime around it.