PickWise workflow

Before you start

• API key from the portal → API keys
• Your work email — the same address you use to sign in to the portal
• MCP in your IDE — follow IDE setup for config templates

Keep your API key in IDE config only — never paste it into chat or task prompts.

At a glance

1. Connect PickWise in your IDE once (API key + your portal email).
2. Begin the ticket you are working on — the standard loads automatically.
3. Code. No PickWise calls needed while you implement.
4. Optional: save a private task note before pausing (max 3000 characters) — restored on Begin task.
5. Log the session when you are done — decisions, recommendations, summary.
6. Optional: add QA, PM, or other follow-up summaries in separate logs.

Tool details: Tools

Steps

1. 1

   Connect PickWise

   Add PickWise to your IDE once. Your admin gives you an API key; use the same email you use to sign in to the portal.

   IDE setup guides

2. 2

   Begin your task

   Tell your agent which ticket you are on. PickWise loads the published standard, decisions, and task context.

   Optional first step: list your tasks if you need names and descriptions before choosing.

   Tools: pickwise_list_my_tasks, pickwise_begin_task

3. 3

   Code with the standard

   The agent applies your task, project, and org standards while coding. Optional: ask PickWise when you need org-wide context beyond the task standard.

   Task notes are private resume scratch space (max 3000 characters) — not Sessions evidence and not shown in the portal.

   Tools: pickwise_save_task_note, pickwise_ask

4. 4

   Log the session

   Log which decisions were applied, any standard gaps as recommendations, and a short summary.

   By default your summary can be posted to the linked ticket. Turn off ticket sync if it should stay in PickWise only.

   Tools: pickwise_log_session

5. 5

   Review in Sessions

   In the portal, open Sessions to see your runs — decisions, notes, recommendations, and summaries.

   Open portal → Sessions

Optional follow-up summaries

Separate logs for QA, PM, or other audiences — same task, each with its own summary.

1. 1

   Standard recommendations

   Note anything you had to guess because the standard was unclear.

   How to log

   Submit each gap as a recommendation on your main standards close-out (pickwise_log_session, session_kind standards).

   Copy for your agent

   Before closing the session, review the task standard for gaps.
   
   For each gap where you had to assume or work around missing guidance:
   - What you assumed or did
   - Why the current standard was unclear or silent
   - What decision should be added so the next developer does not guess

2. 2

   QA handoff

   What was delivered, what tests already cover, and what QA should verify manually.

   How to log

   Log as a separate session: session_kind qa_handoff.

   Copy for your agent

   Write a QA handoff section for the session summary.
   
   Cover:
   1. What was delivered (behaviour, endpoints, UI, config)
   2. What is already verified by automated tests — be specific
   3. What QA should focus on manually: edge cases, auth, validation, errors, and anything tests do not cover

3. 3

   PM assurance

   Complete vs partial vs not done — with a clear readiness call.

   How to log

   Log as a separate session: session_kind pm_assurance.

   Copy for your agent

   Write a PM assurance section focused on feature completeness.
   
   State clearly: complete, partial, and not done — with a readiness line (Ready for QA, Ready for demo, or Not yet ready for production).

4. 4

   Perspective summary

   End user, competitor, and ops lenses on what you shipped.

   How to log

   Log as a separate session: session_kind perspective.

   Copy for your agent

   Write a perspective summary with three sections:
   
   **End user** — What would a real user notice? Bullets + Verdict: one line.
   **Competitor** — How does this compare to strong products? Bullets + Verdict: one line.
   **Ops / on-call** — Debuggability, alerts, failure modes. Bullets + Verdict: one line.

5. 5

   Prompt feedback

   Optional rating of how useful the task prompt was — separate from your main log.

   How to log

   Log as a separate session: session_kind prompt_feedback.

   Copy for your agent

   Rate how effective the task prompt was.
   
   Include Rating 1–10, **What went well**, and **What to improve**.