Guide 05: Read and Write FigJam Boards for Automated Workshop Synthesis

After a workshop, have Claude pull every sticky note and connection off the board, synthesize themes, and write the affinity map back onto the board, with sections, stickies, and connectors included.

Difficulty: Intermediate · Time: 30 min setup, then ~15 min per workshop You'll need: Claude and a Figma MCP server with FigJam tools (read board contents, create stickies/sections/connectors), plus a FigJam board with real content

What changes

Workshop synthesis is the classic "two hours of unglamorous work after the fun part": reading 150 stickies, clustering them, naming themes, writing the summary. If your MCP server exposes FigJam tools, Claude can do the mechanical layer end to end:

• Read: pull every sticky, shape, and text node off a board, plus the connector graph (what's linked to what)
• Synthesize: cluster by theme, surface tensions and outliers, draft insight statements
• Write back: create sections for each theme, generate summary stickies, draw connectors between related clusters, and arrange everything in a tidy grid

The result isn't just a summary in chat; it's a new, organized region on the board itself, where your team already works.

What you need

Connect a Figma MCP server with FigJam tools. In Claude, go to Settings → Connectors (on Team/Enterprise plans, an Owner adds it under Organization settings → Connectors first), click Add custom connector, paste the server's URL, and complete any sign-in it requests. Then, in a chat, click the + button → Connectors and toggle it on for the conversation. Check the server's tool list for FigJam-specific capabilities: reading board contents and creating stickies/sections are the essentials, while connectors and auto-arrange are nice extras. Write access to boards typically comes from community servers that pair with a small desktop bridge plugin you run inside Figma; those need the FigJam board open and the plugin running, and, since a write-capable server can modify your files, vet any third-party server for trust before connecting it.

One prep habit pays off enormously: during the workshop, keep raw input in one region of the board. Synthesis lands in a separate region, so the original data stays intact and auditable.

First run

1. Pull the board. Open the board, then prompt: "Read everything on my open FigJam board. Give me a count by type (stickies, shapes, sections) and tell me if stickies carry author or color information." Sticky color often encodes meaning from the workshop (e.g., pink = pain point); tell Claude your color code if you used one.
2. First-pass clustering. Prompt: "Cluster the stickies into themes. For each theme: a name, a one-sentence insight written as a finding (not a category label), the sticky count, and 3 representative verbatims. Also list stickies that don't fit any cluster." Insist on the outliers list; that's where the interesting stuff hides.
3. Pressure-test the synthesis. This is your judgment step, and it's where you earn your facilitator fee: "Theme 3 feels like two different things; split it and re-sort." or "What's the strongest counter-reading of this data?" Iterate until the themes match your read of the room.
4. Write it back. Prompt: "Create a new region to the right of the raw data titled 'Synthesis'. For each theme, create a section containing a heading sticky with the insight statement and the representative verbatims as stickies beneath it. Use a neutral color so synthesis is visually distinct from raw input."
5. Map relationships. Prompt: "Draw connectors between theme sections that are causally related, labeled with the relationship (e.g., 'drives', 'blocks')."
6. Generate the leave-behind. Prompt: "Now write the workshop summary doc: context, method, themes with evidence counts, tensions, open questions, and recommended next steps." If Claude's file-creation feature is enabled (Settings → Features), have it produce this as a downloadable document for the people who weren't in the room.

Example prompts

Read my open FigJam board. Pink stickies are pain points, yellow
are ideas, blue are questions. Cluster pain points only, and tell
me which ideas map to which pain-point clusters.

Compare this board against the board from our last session
[link]: which themes recurred, which are new, and which pain
points from last time didn't come up at all?

Create a 2x2 on the board (effort vs. impact) as four sections,
and place a copy of each idea sticky into the quadrant you'd
assess it into. Mark any you're unsure about with an orange dot.

Troubleshooting

Claude reads the board but placement is messy. Ask for a grid arrangement with explicit spacing, or use the server's auto-arrange tool if it has one. Give a starting coordinate ("place the synthesis region starting at x=5000 so it's clear of existing content").

Sticky text comes back garbled or truncated. Very dense boards can exceed what one read returns. Read section by section, or ask Claude to paginate.

The clustering feels generic. Feed context first: the workshop's guiding question, who attended, what decisions hang on the output. Generic input produces generic themes.

Write tools missing. Some servers are read-only for FigJam. You can still do steps 1–3 and 6 (read + synthesize + doc) and paste the synthesis back manually; that's most of the value.

Where this takes you

Template the whole flow: save your exact prompt sequence (with your color-coding conventions) as a shared document or reusable Claude instruction set so every facilitator on the team runs synthesis identically. For recurring research programs, keep a running "meta-board" where Claude appends each session's themes; longitudinal patterns emerge that no single workshop shows.

Portability

This workflow is not Claude-exclusive. It is built on the Model Context Protocol (MCP), an open standard, and the same Figma MCP servers described above work with any MCP-capable client, including Gemini CLI, ChatGPT, Cursor, and others. The example prompts in this guide transfer verbatim; only the connector setup steps differ, so follow your client's documentation for adding an MCP server. Two caveats: MCP support varies by client surface and changes quickly, so verify your client supports remote servers (and desktop-bridge pairing, if your chosen server uses one); and model quality shows most on the longer multi-step operations, so expect to supervise more closely and break work into smaller steps with lighter models.