Appearance
Core concepts
This page defines the building blocks of Liddr. Each concept links onward to a deeper guide.
Knowledge graph
The knowledge graph is Liddr's shared model of your organization's reality, assembled from the sources you connect. It holds the structure, work, and knowledge that flow out of your tools — issues, pull requests, wiki pages, incidents, documents — and the relationships between them.
The graph is shared across everyone in your workspace. It has no per-user access control, which is the principle that shapes what may and may not be brought in (see Org-shared vs per-user data).
See The knowledge graph.
Entities
An entity is one item from a source: a Jira issue, a GitHub pull request, a Confluence page, an incident, a document. Each entity keeps its title, its content, and a link back to the original, so every answer can point to exactly where it came from. Entities and the relationships between them are what make up the knowledge graph.
Semantic index
When Liddr brings in a source, it builds a semantic index of the content — an index that lets you search by meaning rather than by exact keywords. Ask a question in your own words and Liddr finds the most relevant material even when none of your wording appears in the source, then traces each answer back to the entities behind it.
Sources & connectors
A source is one of your tools. A connector is the integration that brings that source into Liddr. There are two kinds, and the distinction is fundamental:
- Ingest connectors bring org-shared data into the knowledge graph. Examples include Jira, GitHub, Confluence, Asana, GitLab, Linear, Notion, Sentry, and PagerDuty.
- MCP connectors are per-user integrations. They are not brought into the graph; instead they are consulted live, with each user's own credentials, when an answer needs them. Presets include Slack, Box, Dropbox, Zoom, DocuSign, Gmail, Google Drive, and Google Calendar.
See Connectors overview, Ingest connectors, and MCP connectors.
Org-shared vs per-user data
This is the access principle that governs the whole product. The organization knowledge graph is shared and has no per-user access control, so anything brought into it is effectively visible to anyone who can Ask. From that, two rules follow:
- Only org-shared, non-sensitive data enters the graph — structure, business rules, project tracking, documentation, and team knowledge.
- Personal, record-level, or sensitive data never enters the graph. It is reached through per-user MCP connectors, queried live and scoped to the individual.
This is why communication and personal stores (mail, files, calendars, chat) are reached through MCP rather than brought into the graph: they belong to a person, not to the shared graph.
See the Access model.
Know
Know answers natural-language questions over the knowledge graph and cites the source entities behind every claim. It supports three answer modes:
- Thorough (default) — a long-form answer with structure, tables, and inline citations.
- Concise — a short two-to-four sentence reply.
- Email — a ready-to-send draft for email or chat.
Answers stream as they are written, surface citations as references you can open, suggest follow-up questions, and flag unverified claims that were not found in the cited sources.
See Know.
Signal
Signal is a role-aware intelligence digest. Liddr scores recent activity across your sources, clusters related events, and produces a short narrative — typically a summary line plus five to eight items covering what changed, what's blocked or at risk, decisions, and recommendations. Each item links to its source. You can choose a perspective (for example developer, SRE, architect, PM, designer, or delivery manager) and a time window.
See Signal.
Forge & stages
An initiative is a unit of work that Liddr drives through SDLC stages in order:
discovery → requirements → design → architecture → implementation → review → doneMost stages have an approval gate and produce an artifact before they can be approved, so an initiative advances deliberately rather than all at once.
See Forge.
Build
Build is the implementation stage of an initiative. Liddr drafts the code change in isolation, lets you review the full diff, and — on your approval — opens it as a pull request against the target repository (as a draft by default). You review the change before it is applied and the resulting pull request afterward, so nothing lands without sign-off.
See Forge.
Workspaces
Your workspace is your organization's secure, isolated space in Liddr — your sources, your knowledge graph, and your team, kept separate from every other customer.
See Multi-tenancy.