Plan & track from Claude Code

Use Claude Code (or any MCP-aware client like Claude Desktop / Cursor) to propose plans, push work items, and update lifecycle status directly against your FastLinkIt Projects board — without leaving your editor.

This page is the in-manual summary. The full hands-on walkthrough with conversation transcripts and troubleshooting lives in docs/ProjectPlannerGuide.md in the repo.

What you can do

This is distinct from in-app AI agents (see AI agents — planning & execution). That page covers AI running server-side via the orchestrator. This page covers external AI clients calling the project from outside the FastLinkIt stack.

Two distribution channels

Both surfaces hit the same six API endpoints — pick whichever fits your tools.

Quick start

1. Generate an API key at /Account/Manage/ApiKeys with the projects scope ticked. Copy the raw key (fli_...) — shown only once.
2. Install your chosen surface — see the install steps in Plugins/claude-skill/README.md or Plugins/mcp-server/README.md in the repo.
3. Configure the env vars (FLNKIT_API_KEY required; FLNKIT_API_BASE defaults to https://flnk.it).
4. Test the connection — open a Claude Code session in any repo and ask: "What FastLinkIt projects do I have access to?"

The three moves

Plan

Ask Claude "what's next for project X?" and Claude reads CLAUDE.md + PROJECTS.md, fetches the project's methodology + columns, structures the plan to match the methodology, and surfaces a ranked list with rationale before pushing.

Default plan shape per methodology:

What the proposal looks like before you confirm:

For an Agile project, Claude indents the tree:

1. EPIC   Customer onboarding overhaul                    (no estimate)
2.   FEATURE   Sign-up flow refresh                        (likely: 16h)
3.     STORY   As a new user I want to sign up via Google  (likely: 6h)
4.   FEATURE   Welcome email sequence                      (likely: 12h)
5.     STORY   As a new user I want a 3-email welcome drip (likely: 4h)
6.     TASK    Wire welcome drip into MailingSequenceService (likely: 4h)

For a Waterfall project, items are numbered with a "depends on" column:

#  Type  Title                                  Likely  Depends on
1  TASK  Schema audit + migration plan          12h     -
2  TASK  Stand up replica DB                    8h      #1
3  TASK  Backfill replica from prod snapshot    16h     #2
4  TASK  Cutover dry-run                        6h      #3
5  TASK  Production cutover window              4h      #4

For a Kanban project, a flat priority-ranked table:

#  Type  Priority  Title                                      Likely
1  TASK  high      WIP-limit enforcement on board columns     4h
2  TASK  normal    Assignee avatars on Kanban cards           2h
3  BUG   normal    Drag handle hint on hover                  1h

Override the shape when it makes sense:

• "Don't bother with an epic, this is one cohesive change — just stories." (Agile → flat)
• "Group these into two epics rather than one." (Agile → multiple epics)
• "Don't add dependencies — these can run in parallel." (Waterfall → no chain)
• "Group these under a parent task so they show together." (Kanban → grouped)

Claude reshapes and re-prints the plan; you confirm only when the shape is right.

Always waits for your explicit "yes" before POSTing — ambiguous responses are not consent. On confirmation, items land with the hierarchy + estimates + dependencies preserved.

Pick up

Say "I'll start on FLNKA-21" and Claude opens an agent run server-side. The item moves to the project's in-progress column; anyone watching the board sees the move in real time. Claude stores the run id for the rest of the session.

Wrap up

When the work is done, summarise what changed: "Done — added the WIP-limit check, blocked drops over the limit, tests pass." Claude calls complete_task with the narrative; the item moves to Review (or Done if you say "skip review"). On failure, "Stuck — need design input on whether to block or warn" moves the item to Blocked with the reason logged. Reasons are mandatory — no silent failures.

How it works under the hood

Both surfaces talk to the same six REST endpoints under /api/projects:

The server validates work-item types against the project's methodology — Kanban projects only accept task / bug; Agile projects accept the full hierarchy. Lifecycle transitions reuse the same column-resolution helpers as the in-app orchestrator, so a Claude Code session moving an item to In progress lands in the same column as the server-side agent layer.

What it deliberately won't do

• Push without your "yes". "Looks good" / silence are not consent.
• Pick the project for you when multiple match. Claude lists candidates and asks.
• Modify or delete existing items. Only creates new items + transitions status. Use the web UI for edits.
• Auto-merge code. Code execution is a separate scoping. Today: you write the code, the skill tracks the work.
• Bypass access controls. Same scope and ownership rules as the web UI.

Related

• Project management
• AI agents — planning & execution — the in-app agent path
• REST API — the underlying API both surfaces use