
OpenRouter. Choose what powers the work. Connect your account to access models from multiple providers. Choose what powers each Agent and the embeddings that make your Knowledge searchable. Read the OpenRouter documentation.
Laravel AI SDK. Extend your platform. Choose your infrastructure. Connect privately hosted models, add providers and build custom Agent capabilities on Dollie’s Laravel AI foundation. For Self Hosted and private enterprise deployments, we configure the setup around your infrastructure and data requirements. Explore Laravel AI.
On Enterprise, connect your provider account and control model access across Workspaces. On Cloud, choose from the models available with your plan. Compare model access or explore your own platform.
Browse models through OpenRouter and choose the ones you want to make available. You own the selection, from a focused set to a wider choice for your teams.
Give different customers or departments the model access they need. Keep selection with the platform team or allow Workspace Operators to choose for their Agents.
Give routine intake and summaries a fast, lower-cost model. Choose stronger reasoning for analysis and complex reviews, then evaluate the results on your own work.
Handle frequent requests with a lower-cost model. Configure an intake Agent to read incoming briefs, extract the requirements and create a Task on your board. Your team can review the brief and assign the research to an Agent with a stronger reasoning model.
Give complex decisions a stronger reasoning model. Assign the supplier comparison to a research Agent with the model and tools the job needs. It can compare the proposals against your criteria, investigate gaps and prepare a recommendation for your team to review.
Turn scanned pages into evidence your Agents can use. Index technical PDFs and images with a multimodal embedding model such as Gemini Embedding 2. An Agent can retrieve the relevant diagrams and tables, inspect them with a model that accepts images, and prepare a visual presentation with source references.
Inspect requests with Langfuse Review traced model inputs and outputs to investigate a slow or unexpected answer.
Compare models in use See token usage, latency and model costs for the traffic your deployment traces.
Understand customer economics Use Dollie’s separate usage accounting to understand what each Workspace consumes.
Choose where monitoring lives Connect Langfuse Cloud or a self-hosted instance. Configure OpenRouter Broadcast to send the traces.
Choose the models that power your Agents, set platform defaults, tailor model access across Workspaces and select embedding models for your Knowledge. OpenRouter provides access to multiple model providers through one connection.
On Cloud, Dollie manages the standard model connection. Enterprise uses your own provider accounts, including OpenRouter for access to multiple model providers through one account.
Yes. Choose a fast, lower-cost model for routine requests and a model with stronger reasoning for demanding work. Assign models per Agent in Agent Studio within the access you configure for the platform and Workspace.
Yes. Workspaces separate customers and departments. The platform controls available models and defaults, and can delegate model selection to Workspace Operators. Your teams can have different choices within the same Agent OS.
Its instructions, Knowledge and Workspace permissions remain in Dollie. Its answers, tool behaviour, speed and cost can change, so evaluate the new model on representative work before rolling it out.
Embedding models make source material searchable. Dollie supports text processing and direct-file processing with supported multimodal models for PDF pages and images. A compatible Agent model can receive the retrieved media itself; text-only models use the available text representation. Choose the embedding model per Knowledge Base and reindex when changing it.
Inspect the model requests and responses sent to Langfuse, along with token usage, latency and model costs. The standard OpenRouter setup uses its Broadcast integration to forward traces. Dollie’s usage accounting and customer billing remain separate.
Yes. Run Langfuse on your infrastructure and connect it to Dollie for model tracing and monitoring. For a fully private deployment, we configure the model endpoints, embeddings and monitoring together.
Yes. Private model endpoints are supported as part of a Self Hosted enterprise deployment. We use Dollie’s Laravel AI foundation to connect compatible providers and configure generation, embeddings and Langfuse tracing around your infrastructure requirements. Discuss the setup with us.
It keeps the Dollie application and its stored data on your infrastructure. With OpenRouter, the context needed for model and embedding requests goes to the external service. Keeping all AI processing internal requires private generation, embedding and monitoring endpoints configured for that deployment.
Yes. Business supports a client service on Dollie Cloud. Enterprise adds your own platform brand and domain, provider accounts and model policy. Stripe connects subscriptions and customer billing to the service.
Planning the AI setup for your platform? Discuss your model setup

Run your Agent OS with the models and infrastructure your organisation requires. Start with OpenRouter, or work with us to connect privately hosted models and self-hosted Langfuse monitoring.
Private model endpoints and on-premises monitoring are supported enterprise deployment options. We configure the complete setup around your organisation’s data and infrastructure requirements.