TKCOREAI

Product guide · Projects

TKCORE AI Projects: playbook and benefits for knowledge workspaces (2026)

TKCORE editorial 9 min read

Projects are TKCORE AI workspaces built around one topic, a bundle of source material, and a stable way of chatting. Think of them as a small “reading room plus chat desk” per customer, product line, or internal function—so the model stays aligned with your context instead of mixing everything into one boundary-less thread. For a public overview, see the Projects intro; when signed in, open the library from /projects.

Concept art: multiple project workspace tiles flowing into an AI chat bubble and document stack; deep navy background, indigo accents, flat vector style
Illustrative graphic: “partitioned projects → shared chat and knowledge flow.” The live UI is always the in-app Projects experience.

What problem do Projects solve?

Teams often hit context drift: last week you were on Product A specs, but this week the model still assumes A when you draft Product B email. Or files live in personal folders and everyone pastes ad hoc. Projects add a layer that binds name, topic, description, and writing preferences (tone, length, notes) with documents / knowledge-base material, so one business line keeps a consistent voice across sessions.

Recommended usage patterns

1) One project per business line, account, or campaign

Create a separate project for each major customer, product line, or initiative from the project list. Use a clear project name and topic; spell out guardrails in the description when needed (for example: “2026 pricing only—do not infer unreleased features”). Maintain sources inside the project detail view and keep that project selected while chatting to reduce cross-talk.

2) Documents first: ingest, index, then ask

Upload or attach PDFs, Markdown, plain text, and similar formats inside the project workspace (supported formats and indexing states depend on your backend). A practical rhythm is: upload → wait for indexing → sanity-check with short questions (for example: “From section three, list three compliance caveats”), then move into drafting or customer-facing answers. You can still pair with specialised tools from the tools directory, but treat project context as the primary anchor for facts.

3) Home chat and the ?project= query

When you jump from a project tile into home chat, the route carries a project query string (for example ?project=prj_abc123) so send/save-style APIs line up with the active workspace. That matters when you switch projects in the sidebar or share a bookmark: the URL and the active project stay in agreement, lowering the risk of “UI shows B while the request still targets A.”

4) Signed out vs signed in: local-first vs cloud-backed

On the browser, the project library is cached locally as a fallback. With a valid signed-in session (JWT), list/detail CRUD and document flows go through /api/projects so the same project library can follow you across devices. If you are not signed in, you can still learn the “project” organising pattern locally; for team-grade persistence and KB indexing, sign in and prefer server-backed prj_… project IDs when calling chat APIs.

From the project library to project-scoped chat

TKCORE Projects flow: project list, project detail workspace, home chat with project query parameter /projects List · create · delete Open detail Workspace Sources · notes · prefs In-app chat / toolbar Home chat ?project=prj_… Aligned context
Conceptual flow: manage entries in the project list → curate sources and preferences inside a project → when needed, use home chat with a project query parameter to lock the same knowledge boundary.

Why Projects help

  • Isolation: workspaces do not bleed into each other—ideal for multi-customer, multi-regulatory, or multi-locale brand lines.
  • Sources + preferences together: topic, description, and tone/length notes live with the project, so you repeat fewer “system prompt” pastes.
  • Routing matches chat: resolving ?project= from the address bar first reduces mismatches right after you switch projects.
  • Progressive adoption: learn the workspace locally, then sync through /api/projects once signed in—solo trials and team rollouts share one mental model.
  • Works with the tool directory: Projects anchor facts and boundaries; specialised tools handle format and genre—closer to how product and ops teams actually ship.

Planning for capacity

The browser-side project library enforces limits on total projects and documents per project to protect performance and storage quotas; cloud-backed projects follow server rules. If you approach the cap, merge closed initiatives into archive-style projects, or keep long-tail sources as external links with only the critical excerpts ingested.

Suggested next steps

  1. Read the Projects intro and study the sample layout.
  2. After signing in, create two or three projects you will actually reuse—not infinite fragmentation.
  3. Seed each project with three to five canonical documents you cite every week; validate Q&A and drafting quality before scaling uploads.
  4. When you need polished outbound copy, pair with the right runner from the AI tools directory for tone and layout passes.

Related: Projects list, Projects overview, FAQ, Contact.

More articles