JotGroveNOTES · PLANS · REVIEW

How AI works · more than MCP

AI can clarify and plan.
It cannot decide in your place.

JotGrove uses small specialist Agent roles for notes, decisions, projects, risks, and reflection. It also suggests tags, helps organize selected Jots, can work from a confirmed Reset Your Life blueprint, and lets external AI clients use a limited MCP connection.

Open JotGrove
At a glance
  • Choose 1–3 of six specialist Agent roles
  • From Jots: one source plus up to 20 selected context Jots
  • Every actionable result stays a draft first
A typical AI run

I want a calmer start to the week

  1. NOTE
    Source

    The note stays unchanged; server-backed runs bind its revision.

  2. TEAM
    Agents

    Choose roles or use the team suggested for this kind of note.

  3. DRAFT
    Artifacts

    Review clarification, options, a plan, tasks, risks, or reflection separately.

  4. YOU
    Decision

    Accept, reject, or leave each result unchanged.

Source note: unchanged · Approval: required

The operating rule

AI work is visible, scoped, and reversible until you approve it.

An AI run is attached to one source note and records the roles and drafts involved. Provider-backed runs also bind the source revision so the server can reject an outdated result. It is a review process—not an invisible background rewrite.

On this page
  1. 01AI Agent team
  2. 02Context control
  3. 03Drafts and approval
  4. 04AI tags
  5. 05Jots and Reset Your Life
  6. 06AI Connections · MCP
01

AI Agent team

Use the roles that fit the note instead of one generic assistant.

JotGrove can suggest a team from the note type, and you choose one to three roles for the run. The available roles are Clarifier, Decision Analyst, Planner, Implementation Guide, Risk Reviewer, and Reflection Guide.

  • Clarify an ambiguous direction or missing detail
  • Compare options, break down a plan, and prepare implementation steps
  • Surface risks or reflect without forcing a plan
02

Context control

A single Card is the default. Extra context currently starts in Jots.

Starting AI from a Wall Card, Jot, or Note detail uses that Card alone. To add context today, multi-select at least two Jots: the first becomes the source and up to twenty more become context. The detail view has no arbitrary related-note picker. The Web app does not scan the whole Wall and sends AI runs with online research disabled.

  • Make the multi-selection visibly in Jots before confirming the Agent team
  • Keep every unselected Card outside the prompt
  • Provider credentials remain on the server, never in the browser
03

Drafts and approval

Review each output as a separate artifact.

A run may return a summary, clarification questions, decision options, a plan draft, task drafts, implementation notes, a risk review, or a reflection. Analysis can be marked useful or dismissed; plan and task drafts are accepted together.

  • The original note is never overwritten
  • Plan and task records are created only after approval
  • Provider-backed drafts check the source revision again when applied
04

AI tags

Ask for tag suggestions while editing a note.

AI can use the note title, body, current tags, and available workspace tags to propose a short list. The suggestions appear in the tag editor, where you decide what to add.

  • Suggestions do not replace existing tags
  • Duplicate and invalid tag names are filtered
  • You can still create, rename, color, or remove tags manually
05

Jots and Reset Your Life

AI can enter from more than one part of the workspace.

Select at least two Jots when you want one AI run to analyze them together. The first becomes the source and the rest become explicit context; the run does not immediately merge, tag, or move them. In Reset Your Life, confirm your own responses before passing the blueprint into the ordinary AI draft flow.

  • Multi-Jot analysis starts only from the items you select
  • Reset answers remain editable and are not inferred by AI
  • Any plan or task produced later still follows the draft review flow
06

AI Connections · MCP

Let a supported external AI client read a limited synced view.

After OAuth 2.1 consent with PKCE, search, fetch, and list_recent read synced Cards as note resources, plus synced plans and tasks; list_tags separately aggregates tags from synced Cards. submit_draft accepts note, knowledge_summary, or plan_outline drafts and sends them to JotGrove for review.

  • MCP sees only content already synced to the server
  • Approving a submitted draft creates a Note or Jot—never a Plan—and cannot edit or delete existing content
  • Revoke one connection or all connections from Settings

Clear boundaries

What AI does not do in the current Web app.

The useful part of AI here is a controlled draft workflow, not unattended automation.

  1. 01

    Off, suggest, and auto control when runs begin; even auto results still wait for review

  2. 02

    It does not read the whole workspace, research the web, overwrite a note, or silently create a plan

  3. 03

    Real model runs and AI tag generation depend on the server being configured and reachable

Try a bounded run

Start with one note and one question.

Choose the roles and context, inspect every draft, and apply only what is actually useful.

Open JotGrove