LinkContext

LinkContext is a team's shared store of LinkedIn facts, kept beside the CRM: profiles, companies, career history, connections, invitations, posts, interactions, notifications, message threads, and bookmarks. The browser reader is read-only. An assistant ingests and updates through the linkcontext client.

One deployment is one team's workspace. Records are ordinary database rows the team can read, not client-encrypted files the way VaultContext stores them. Until a public site is configured, the application listens on port 8095 and can bootstrap one local user. NetworkContext uses that same default port and does not share this database.

The README's purpose for the data is cross-reference with DealContext. After a person checks a match, profiles can store crm_person_id, crm_organization_id, crm_deal_id, and crm_message_id. Those are links the team writes. They are not a live feed from the CRM.

What is stored

flowchart TB
  profile[Profile]
  company[Company]
  career[Career history]
  conn[Connections and invitations]
  post[Posts and interactions]
  thread[Message threads]
  bookmark[Bookmarks]
  crm[DealContext, as a neighbor]

  profile --> career
  profile --> company
  profile --> conn
  profile --> post
  profile --> thread
  profile --> bookmark
  profile -.-> crm

Writes carry expected_revision. If someone else saved first, the write conflicts and the assistant re-reads. Deletes through the API are blocked. A record's lifecycle status changes in place, and an audit row records the change. Upserts are idempotent: the same LinkedIn identity maps to the same record id, so a second sync updates instead of duplicating.

sequenceDiagram
  participant Agent
  participant LC as LinkContext

  Agent->>LC: sync-start --scope feed --since 7d
  Agent->>LC: upsert profiles and posts
  Agent->>LC: sync-finish with counts

LinkContext stores the team's curated LinkedIn context. It is the shared store. The local, one-account capture archive is NetworkContext.