An open-source, AI-first support platform: a knowledge base plus a conversational agent that answers from your articles with sources attached, and hands low-confidence conversations to a human inbox. MIT licensed, self-hostable.
RETRIEVEANSWERESCALATE
no signal is not a low signal
Each of these exists in the codebase. We name the mechanism so you can hold us to it.
Answers are retrieved from your knowledge base via embeddings and vector search, and every AI reply cites the articles it used. Grounding you can click.
Each answer carries a grounding-based confidence score checked against a per-workspace threshold; below it, the conversation escalates to the human inbox. Absent confidence is treated as "no signal", not "low": the gate is deliberate, not a naive number compare.
Questions your KB cannot answer are collected and ranked by frequency, with one-click article creation. The gaps become the roadmap.
Anthropic, OpenAI, Google, or Mistral per workspace with your own API keys, encrypted at rest with AES-256-GCM.
One script tag embeds chat and search with session persistence and a "Talk to a human" button.
Your knowledge base speaks three machine protocols: llms.txt, an MCP server (search, read, ask), and Google’s A2A. Support for AI customers, not just human ones.
Bootstrap a KB from your codebase, PRs, and releases. Everything lands as a draft flagged AI-generated. A human publishes.
Eight built-in themes plus migration CLIs from Intercom, Mintlify, and Notion.
The SDK, MCP server, themes, and integrations ship as public npm packages under @helpnest.
Where it standsOpen source under MIT: the complete self-hosted app is on GitHub. A hosted cloud version exists as a separate closed service; the open-source repo is the real product, not a teaser.
AI phone receptionist that answers calls and books the appointment.
MCP gateway that gates every agent tool call.
Local-first desktop orchestrator for coding-agent CLIs.