Named AI agents join your team chat, read your documentation, and become your team's knowledge expert.
Three problems every engineering team faces.
Wiki pages written once, never updated. New hires read outdated runbooks. Nobody knows which version to trust.
RAG retrieves text chunks. It doesn't know what your team does, which docs conflict, or what was validated last week.
When senior engineers leave, their knowledge goes with them. There's no system that captures and preserves what the team knows.
Harry is a named LioraAgent that lives in your team chat — Slack, Teams, or wherever your team communicates. He reports to a manager, reads your docs, and answers questions with structured, cited knowledge.
Point Harry at your wiki — Confluence, Notion, SharePoint, or a plain URL. He crawls it, extracts key steps, facts, and owners using LLM synthesis. Not raw text — structured knowledge.
Entries are linked: references, depends_on, derived_from, contradicts. Browse connections in an interactive D3 graph.
Harry only talks to people his manager approves. Unknown person? He creates a group DM with the manager to ask permission.
Two docs say different things? Harry detects it, notifies the manager, and asks "which one should I trust?"
Other AI agents connect to Liora's knowledge via MCP Server or REST API. Your on-call agent, code review bot, or postmortem writer can all query and contribute knowledge.
9 tools: search, get, ingest, challenge, memory, graph, changelog. Any MCP-compatible agent connects natively — Claude, GPT, custom agents.
120+ endpoints. Register agents with scoped API keys (read, write, challenge). Rate-limited, tenant-isolated, audit-logged.
Subscribe to knowledge changes via WebSocket or Kafka. Get notified when docs update, contradictions are detected, or new knowledge is added.
Everything you need to manage, explore, and govern your knowledge base.
Wiki-style browsing with semantic + full-text search. Every entry shows product, owner, key steps, key facts, source URL, validator, and confidence score.
Interactive D3 visualization. 6 relationship types. BFS traversal. Click through connections. Impact analysis. Reading order from depends_on chains.
Durable facts Harry learns from documents and conversations. "We use ArgoCD for deployments." "auth-gateway is the critical edge service." Searchable, citable.
Every LLM call logged: prompt, response, tokens, cost, latency. See exactly what goes to external AI providers. PII redacted before sending.
Side-by-side comparison of conflicting entries. Resolve with explanation. Resolution history. LLM-confirmed severity (critical/high/medium/low).
Confluence, Slack, Teams, SharePoint, Google Docs, Notion, GitHub. Credentials encrypted in database. Test connectivity from the UI.
Knowledge sources, messaging platforms, AI providers, and databases.
| Traditional RAG | LioraEngine | |
|---|---|---|
| Knowledge | Retrieved at query time, forgotten after | Accumulated over time, structured, versioned, linked |
| Understanding | Text chunks with similarity scores | Product, owner, key steps, key facts extracted by LLM |
| Contradictions | Returns both conflicting docs without noticing | Detects conflicts, notifies manager, tracks resolution |
| Sources | No provenance | Every fact traceable to source URL + validator + date |
| Memory | Stateless — starts fresh every query | Persistent agent memory across sessions |
| Access control | Anyone can query | Manager-approved contacts only. Unknown users gated. |
| Other agents | Not designed for agent-to-agent | MCP Server (9 tools) + REST API + event subscriptions |
Deploy as a VM or AMI. Your data stays in your infrastructure. LLM calls are PII-redacted. Full cost visibility.
Request a Demo