Connect your help center or documents, configure how the agent responds, and embed it on your site. Route requests to your team when human follow-up is needed.
For teams answering recurring customer questions with their own help content.
This dialogue is illustrative, not a product screenshot or customer result. The Support tab reflects the initial use case: general business FAQs and support. Workflow and Setup tabs show interaction patterns, not confirmed launch availability.
Uses help-center and support content as context for responses.
How do I reset my password?
An agent can summarize the steps in your help article and show the source it used.
Actual setup screen from an empty local test workspace, before creating an agent. It shows response mode, greeting, and next steps; it does not show a customer conversation or a published agent.
Set up and publish an agent in three steps.
Upload PDFs or Markdown files and connect URLs to prepare content for retrieval.
Choose your model, set the response mode (Knowledge-only or Assistant), and tune the prompt instructions.
Add the widget to your website or use the API to build a custom interface.
Connect shared knowledge, configure responses, and manage access for each team.
Switch between strict "Knowledge Only", "Hybrid", or "Free Assistant" modes depending on each workspace use case.
Usage-based billing with an append-only ledger. Track exactly which agent and which mode is consuming your resources.
Capture leads and human help requests when AI isn't enough. Integrated status management and ticket claiming.
Adjust how many content sections are retrieved and how they are ranked to provide answer context. Review citations and evaluations to inspect agent behavior.
Define scorecards with "must-include/exclude" rules and run automated tests to evaluate response criteria.
See document freshness and the status of re-indexing work.
Built on Google Cloud infrastructure
Monthly plans start at $24. Compare agent limits, included credits, and team controls.
Monthly prices are in US dollars (USD). No free trial is available; this page collects launch-access requests and choosing a plan does not start a subscription.
Plan credits expire at the end of each billing period. Purchased top-ups carry forward until used; document ingestion can also use credits. See top-up prices and usage details.
For a single workspace or proof of concept.
For teams sharing knowledge across multiple agents.
For production assistants with governance and integrations.
Need custom branding, domains, usage limits, or onboarding? Talk with us about availability.
The differences that matter most for a launch decision.
Eosphorus is a founder-led startup built by Antonio Roba Aparicio, who owns the product architecture, AI/RAG research direction, and Google Cloud-oriented production infrastructure.
Antonio leads the technical build of Eosphorus: tenant-isolated AI assistants, retrieval-augmented generation, RBAC controls, credit ledgering, observability, and cloud deployment for production operations.
See what’s improved in Eosphorus and how the platform continues to evolve.
Loading updates…
Quick answers for technical teams who want the basics first.
For general business FAQs and support. Do not configure agents to solicit health information, children's personal data, or payment card numbers, and do not upload documents containing them. Visitors may still type sensitive information; contact us if that happens or before a higher-risk use case. Card details belong with the payment provider, not in chat.
This form records your interest; it does not grant immediate access or start a subscription. You’ll receive a confirmation email with a link to remove your request. No response time is published. Product updates are optional.
There is no time-limited trial. This page collects launch-access requests; choosing a plan here does not start a subscription. Prices are shown monthly in US dollars.
Credits included with your plan are granted for each billing period and expire at its end. Purchased top-up credits carry forward until used. Document ingestion can also use credits. Review usage by agent or tenant in the credit ledger. Top-up packs cost US$5 for 5,000, US$15 for 20,000, and US$35 for 50,000 credits.
Responses use credits based on generated tokens: 1 per token in knowledge-only mode, 2 in knowledge-plus-assistant mode, and 3 in assistant mode. For example, a 500-token generated answer uses 500, 1,000, or 1,500 credits respectively. Document ingestion uses 10 credits per successfully embedded content chunk. Without your response lengths, mode mix, and source sizes, we cannot estimate how many conversations a plan will cover.
Operations that require credits cannot be processed when the balance is insufficient. You can add a top-up or wait for credits to be granted in the next billing period.
We do not use customer content to train our foundation models. AI providers process content needed to deliver the service; see the privacy policy for providers and details.
Workspace access controls and database row-level security help separate tenant data.
We can discuss custom branding, domains, usage limits, and support. Contact sales about availability; deployment in your own cloud is not currently offered.
Yes. Upgrades take effect immediately. Downgrades apply at end of billing period.
Authorized members can export account and workspace data and erase visitor handoff records. The export does not include chat transcripts. Contact us before using the service with visitor personal data to arrange applicable processing terms.
Google Gemini generates agent responses. OpenAI text-embedding-3-small creates embeddings used to retrieve source content. Providers process the content needed for these functions; see the privacy policy for details.
Fine-tuning changes model weights through training. RAG retrieves relevant content from your documents and provides it as context for a response. This does not guarantee every response is correct.
You can add documents or URLs as knowledge sources. If you need to import agents, settings, or content in bulk, contact us to confirm the available option for your setup.
SSO/SAML availability depends on plan and workspace access. Contact sales to confirm whether your identity provider is supported.
Every credit consumed is recorded in a transparent ledger. You can audit exactly how many credits each agent or tenant is using in real-time.
Connect your business content, configure how the agent responds, and embed it on your site.
