Owlish ingests your knowledge, runs grounded AI agents on top, and gives your support team the inbox, handoff, and analytics to operate them. Six pieces, one console, one bill.
Six things have to be true before an AI agent reliably handles support. Owlish ships all six in the box.
Crawl your website and help center, upload PDFs, DOCX, Markdown and CSVs of resolved tickets. Schedule auto-sync so the agent stays current. OCR fallback for scanned docs.
Retrieval over your indexed content with citations on every answer. The agent will not invent a refund policy, hours, or product spec — if it isn't in your knowledge, it says so.
Let the agent run lookups, book meetings, file tickets, update CRMs, or call your own REST endpoints. Every tool is logged, scoped, and rate-limited per agent.
One agent, every surface. The web widget, Slack, Microsoft Teams, Google Chat, WhatsApp, Messenger, Telegram, the public API, and an MCP server for developer agents.
When the agent hesitates or the question is delicate, it routes to a teammate with the full transcript, the citations it tried, and the visitor context. No re-explaining.
Save a corrected reply as new knowledge in one click. CSAT, escalation reasons, knowledge gaps, and A/B results loop back into the dashboard so the agent gets smarter weekly.
The console is the admin and operations surface for your team. Built like an OpenAI-style developer console — fast, keyboard-first, no-nonsense — with a shared inbox and the controls operators need to ship great support without juggling six tools.
Every conversation across every channel in one queue. Filter, tag, assign, whisper to the agent mid-thread, or take over entirely.
Test prompts and models side-by-side on your live knowledge base before you ship. Compare answers, citations, and latency.
Crawls, uploads, Q&A pairs, and helpdesk imports. See ingestion status, processed size, and which sources back which answers.
Tools and actions the agent can run, with per-skill enablement and rate limits. Bring your own endpoints or pick from built-ins.
Web widget, Slack, Teams, Google Chat, WhatsApp, Messenger, Telegram, MCP server. Connect once, the same agent serves every surface.
Resolution rate, escalation reasons, knowledge gaps, CSAT, A/B results. Weekly insight digest in your inbox.
The widget runs on your site — and on Slack, Teams, the API, anywhere your customers reach for help. Your visitors don't see Owlish; they see your brand, answering them in real time, citing the page they should read next, and routing to a real teammate the moment the agent is unsure.
You can export from Workspace → Data → Export. Choose CSV and the date range, then we'll email you the file.
Same knowledge, same skills, same operator inbox — across the channels your customers already prefer. Live now: web widget, Slack, Teams, Google Chat, WhatsApp, Messenger, Telegram, public API, MCP server. Coming next: Email, Instagram, Discord.
A real public API and an MCP server, not a webhook fig leaf. Your team can ingest, query, and observe Owlish from your own tools.
Workspace-scoped API keys. Ingest sources, drive sessions, query agents, fetch transcripts. Versioned, documented, rate-limited.
Plug Owlish into your own dev agents through Model Context Protocol. The same skills your support agent uses, available to your tooling.
HMAC-signed outbound events on source status changes, conversation lifecycle, escalations. Wire it into anything that listens.
Knowledge and conversations encrypted at rest and in transit. Workspace-scoped storage. Your content is never used to train shared models.
DPA available on request. Built on layered tenant isolation: Auth, Data Connect filters, Server Actions, and DB triggers — defense in depth.
Delete a source, a conversation, or the whole workspace any time. Export sources, transcripts, and config as JSON or CSV whenever you want.
Start Free without a card. Upgrade only when you need more capacity. Usage caps protect you from runaway AI bills.