Pen-first AI canvas landscape: who ships "the agent writes back," and what do they charge
The question
"What are the top pen-first AI canvas/whiteboard tools launched or repositioned since 2024, how are they monetizing, and which user segments are they winning?"
Context: ink.rdco.dev launched 2026-07-21 with a stated productization roadmap. The founder needs competitive context before the second sprint. The live sub-question is whether "pen-first + agent writes back to the canvas" is OWNED, CROWDED, or EMPTY.
What we already know (from the vault)
- ink is already more built than a prototype. Per [[2026-07-21-ink-build-spec]] and [[IMPLEMENTATION-NOTES-2026-07-21-ink-v0]], v0 plus same-day patches (v0.1-v0.3.1) ship: pressure-sensitive pen input with palm rejection, a triple-circle gesture that is simultaneously "send this" and "answer here" (spatial addressing), a fade-absorb animation, page-PNG + region-crop sent as vision input to the Anthropic Messages API, and a reply rendered as animated handwriting (Dancing Script via opentype.js, per-glyph jitter/rotation/slant) plus agent-drawn SVG diagrams resampled into real pen strokes. It is a static PWA, BYOK, deployed behind Cloudflare Access.
- The category claim is already written down, verbatim from the spec: "whiteboard apps have no brain; chat apps have no pen. ink is a pen-first thinking surface where an agent writes back - into the spot you point at."
- Monetization is a direction, not a model. The spec's only monetization language is v2's "Option-4 monetization: RDCO-held tokens, per-user metering, Stripe; Access → real auth for strangers." No price, no target buyer, no sizing. Current scope is explicitly "private: founder + Andrew + Ray."
- One prior art is already acknowledged:
MaximeRivest/riddle(reMarkable Paper Pro; handwrite → page-PNG → vision LLM → reply animated as handwriting). The ink README credits it: "None of its code, all of its magic." No other competitor appears anywhere in the ink docs, which is the gap this brief closes. - The commoditization thesis is relevant and uncomfortable here. [[2026-05-04-karlmehta-llm-commoditization-intelligence-rails]] argues the durable layer is the harness, not the model. Applied to ink: the handwriting-render and the gesture are the durable layer; the Anthropic call is not.
What the web says
The pen-FIRST cluster (stylus is the primary input) monetizes on cheap annual subs, hardware attach, or free-as-platform - and almost none of it writes back.
- Goodnotes 6 (v6 launched 2023-08-09, AI expanded through 2025) is the volume leader in consumer/student pen notes: 4.7 stars across 381,000+ App Store reviews. Pricing fetched from goodnotes.com/pricing: Free (3 files, watermarked exports), Essential $11.99/yr, Pro $35.99/yr, Teams $120/seat/yr, plus an AI Pass add-on at +$9.99/mo for expanded AI credits. Ask Goodnotes is capped at 5 queries/month on free tiers. Writes-back status: partial and narrow - spellcheck and Word Complete render corrections in the user's own handwriting style on the page (per Goodnotes support docs), but Ask Goodnotes, summaries, and Writing Tools all live in a side panel. A Sept 2025 BusinessWire release claims Goodnotes can "generate visuals from text or handwriting... directly in your canvas"; that claim is press-release-sourced only (direct fetch of the release timed out and support.goodnotes.com 403'd), so treat the scope of canvas generation as unverified.
- reMarkable shipped Paper Pro 2024-09-04. Pricing fetched from remarkable.com: Paper Pro from $629, Paper Pro Move from $449, and Connect at $3.99/mo (Connect for Business from $6.99/mo) gating cloud AI, handwriting search, summaries, and Slack/Drive/Miro integrations. Classic hardware-attach: the sub is a cheap annuity on an expensive device. Segment: premium distraction-free deep work (writers, executives); evidence is press-only (TIME Best Inventions 2025), no usage numbers. Writes-back: ASSISTS-ONLY - converts and summarizes, never authors new ink.
- Notability (AI rolled out across 2024-2025): Plus $15.99/yr, Pro $79.99/yr per notability.com/pricing, with tier-gated AI quotas (100 YouTube conversions/mo, 400 quizzes/mo on Plus; unlimited plus "Chat with your Notes" on Pro). Caveat: the same page listed monthly figures ($19.99/$99.99) that are implausibly high relative to the annual prices - the annual figures are the trustworthy ones. Segment: lecture-capture students. Writes-back: ASSISTS-ONLY.
- Apple Notes / Freeform + Apple Intelligence shipped Smart Script and Math Notes with iPadOS 18 (announced WWDC June 2024, shipped Sept 2024). Free. Math Notes is the single clearest incumbent writes-back: it solves handwritten equations and writes the answer inline in a font generated to match your own handwriting. Smart Script rewrites and reflows your ink in place. This is the structural floor of the whole category and the most under-rated competitive fact in this brief.
- MyScript Notes (ex-Nebo) has the best offline recognition engine (4.6 stars,
71.8K reviews) and has bolted on a GPT-backed Summarize/Explain/Quiz layer that renders in a side-panel card UI. Price UNVERIFIED (vendor page carries no dollar figure). Concepts shows no shipped generative AI at all; its Pro sub ($5/mo, $30/yr) is search-sourced and UNVERIFIED (pricing page 404'd on fetch). Supernote, Boox, Kindle Scribe are all hardware-attach and all ASSISTS-ONLY; Kindle Scribe's "script font" re-render is the closest borderline case.
The canvas-first cluster (mouse/keyboard primary) does write back - and monetizes at 3-10x the price of pen apps, per seat.
- Miro (AI Nov 2023; Intelligent Canvas + AI Sidekicks July 2024) - fetched miro.com/pricing: Free; Starter $8/user/mo (25 AI credits/member/mo); Business $20/user/mo (50 credits, full agents-on-canvas); Enterprise custom. Sidekicks act on the board. Segment: large-enterprise cross-functional teams.
- Figma / FigJam AI (FigJam AI Nov 2023; Figma Make GA 2025-07-24) - fetched figma.com/pricing: Starter free (150 AI credits/day); Professional $16/mo, Organization $55/mo, Enterprise $90/mo per full seat annual, AI bundled as credits rather than metered separately. Figma also acquired Weavy (AI-canvas startup founded 2024, $4M seed) in Oct 2025 and rebranded it Figma Weave. Segment: product and design teams.
- tldraw - "Make Real" Nov 2023, tldraw computer 2024, an agent starter kit presented June 2026. Its AI docs describe a typed action schema where validated LLM output is applied directly to the editor instance in real time: full read/write canvas access, not a side chat. Monetization is an SDK license with no published price ("value-based pricing," 100-day trial, free hobby/startup licenses on application). Segment: developers building canvas products - i.e. tldraw is a supplier to this category, not only a rival.
- Excalidraw+: $6/user/mo annual ($7 monthly) per plus.excalidraw.com/pricing, free tier limited to 1 scene with "Limited" generative AI. Text-to-diagram and wireframe-to-code both emit real editable elements onto the canvas. Segment: open-source dev diagramming.
- Muse is gone as a competitor. museapp.com now 301-redirects to allume.com. Per Adam Wiggins' own retrospective the venture-backed team unwound; a solo developer continues it. Allume pricing fetched: Free (2 workspaces, 100 cards), $9.99/mo or $99.99/yr, Setapp bundle $14.99/mo. AI is a passing "AI support with MCP" mention, no agent-on-canvas evidence. This is the single most instructive data point in the brief - the best-funded, best-designed, most explicitly Apple-Pencil-forward "tool for thought" canvas of the last decade failed as a venture business and survived as a $99.99/yr solo product.
- Heptabase has "AI Agent editing whiteboard content" on its public roadmap - announced, not shipped.
The exact wedge: someone shipped it 7 days before ink launched.
- PenEcho (github.com/penecho/penecho) - verified directly against the GitHub API: repo created 2026-07-14, last pushed 2026-07-25, 1,701 stars / 188 forks, AGPL-3.0. Self-described: "Think with AI beyond the chat box. A shared canvas for handwriting, equations, diagrams, and spatial reasoning." README (fetched) confirms the mechanic: you "put a question, equation, diagram, or half-formed idea anywhere on the canvas and pause," the model reads the marks and their spatial relationships, and "answers beside them." Supports Claude Opus 4.8 / Fable 5, GPT-5.6, Kimi K3. AGPL plus "an alternative commercial license... for proprietary products and hosted services." No price published.
- Microsoft Research validated the concept a rung higher: ImaginationVellum (UIST '25) treats the canvas as "an active prompt space" and introduces "Generative Strokes" - AI-generated freeform strokes placed spatially in response to ink. Research prototype, not a product. (ACM/MSR pages returned 403; mechanics are abstract-sourced, medium confidence.)
- Adjacent-but-not-it: OneNote Copilot gained inked-note analysis ~2026-04-29 (panel-driven output). Samsung Note Assist cleans up your own handwriting, free. Cuneflow is a pre-Kickstarter e-ink prototype whose AI output explicitly lands in separate tabs, not on the ink page. The rumored OpenAI/Jony Ive pen is reported screenless, which structurally rules it out of this wedge.
Convergences and contradictions
- Convergence, and it stings: the ink spec's category claim ("whiteboard apps have no brain; chat apps have no pen") was accurate when written and is now roughly 60% accurate. It is still true that no pen-first commercial product has an open-ended agent authoring on the canvas. It is no longer true that nobody ships the mechanic - PenEcho does, publicly, with 1,701 stars in 13 days.
- Contradiction with the ink docs' implied whitespace: the docs name exactly one prior art (
riddle) and no competitors. The real map has at least three live threats the docs don't see - PenEcho on the mechanic, Apple Math Notes on the handwriting-render (free, on every iPad), and Goodnotes on the pen-first install base with an AI-credit SKU already priced. - Convergence on monetization shape: Goodnotes' $35.99/yr base + $9.99/mo AI Pass is precisely the shape of ink's roadmap "option 4" (RDCO-held tokens, per-user metering, Stripe). That is validating for the model and sobering for the price ceiling - the pen-first segment pays ~$12-36/yr for the base product and treats AI as a ~$10/mo add-on. The $8-90/seat/mo pricing lives on the canvas-first, team-collaboration side of the line, not the pen side.
Synthesis for RDCO
Verdict: CROWDED, and it got crowded in the last thirty days. Not owned - nobody has a commercial product, a price, or a business in this exact spot. Not empty - the core mechanic (pen input → agent reasons over ink and its spatial layout → answer placed on the same canvas) ships today in PenEcho, was validated by Microsoft Research a year earlier, and was already prior art in riddle, which ink's own README credits. Calling this whitespace would cost a sprint.
The nearest competitor is PenEcho, and the differences are real but narrower than they feel. Three of them hold up: (1) ink's reply is rendered as handwriting - animated strokes, per-glyph jitter, agent-drawn SVG resampled into pen strokes - where PenEcho returns typeset text, LaTeX, and HTML widgets you accept or discard; (2) ink's trigger is a deliberate spatial gesture, the triple-circle that means "send this and answer here," where PenEcho auto-fires on a configurable pause and supports stylus and mouse equally, making it pen-tolerant rather than pen-first; (3) ink is a hosted web PWA where PenEcho is a local desktop app on npm and localhost:3888. What does not hold up as a differentiator is BYOK, which both do, and which is the thing that caps both at the technical-tinkerer segment.
The segment map explains where the money actually is, and it is not where ink currently sits. Goodnotes and Notability own students. reMarkable owns premium deep-work professionals via hardware attach. Concepts owns sketching designers. Miro owns enterprise teams, Figma owns product teams, tldraw owns the developers who build canvases. Allume - the ex-Muse - owns individual deep thinkers at $99.99/yr as a solo-run business, which is the closest existing analog to ink's plausible buyer and a useful reality check on ceiling. ink today owns the same segment PenEcho owns: AI-curious technical tinkerers who will paste an API key. That is a real segment and a bad business, because BYOK selects for the users least willing to pay for the layer above the key.
Two structural facts should shape sprint two more than the monetization plumbing does. First, Apple is the floor and it is free: Math Notes already writes generated answers back onto the canvas in a font matched to your handwriting, on every iPad, at zero marginal cost. Anything ink does that Apple could fold into Notes is not defensible; the defensible surface is open-ended agent reasoning over an arbitrary canvas with grounding in your corpus - which is exactly the v1 "KB grounding" and "export page → vault note" items, not the v2 billing items. Second, the pen-first audience is overwhelmingly on iPad, and ink is a web PWA. Every player that actually wins a pen segment - Goodnotes, Notability, Concepts, Apple - is native. The roadmap already contains the right answer ("iOS native (PencilKit) reusing the stroke model"), but it is filed under v2 behind live collaboration and Stripe. On this evidence that ordering is backwards: PencilKit is a distribution decision and billing is a monetization decision, and there is nothing to monetize until distribution reaches people who own a Pencil rather than people who own an API key.
The honest strategic read is that ink's differentiated asset is the handwriting-render plus the gesture, not the agent-on-canvas idea, which is now contested. Per [[2026-05-04-karlmehta-llm-commoditization-intelligence-rails]], that is the correct shape - the durable layer is the harness, not the model call. Sprint two should sharpen the two things PenEcho does not do (ink-rendered replies, deliberate spatial addressing), close the grounding loop into the vault, and move PencilKit forward. Pricing can wait; there is a validated anchor sitting at roughly $10/mo for AI credits on a ~$36/yr base whenever it is needed.
Why this is in the vault
This brief is the missing competitive section of [[2026-07-21-ink-build-spec]], whose competitor list contained exactly one entry (riddle) and therefore encoded an "empty niche" assumption into the v1/v2 roadmap ordering. It exists to force one specific decision before ink sprint two: whether to keep PencilKit and Stripe in v2, or to promote iOS-native ahead of monetization now that the web-PWA-plus-BYOK position is known to be shared with a faster-moving open-source competitor.
Open follow-ups
- Who is behind PenEcho, and is it a company or a weekend project? A 13-day-old AGPL repo with a stated commercial-license option and a Moonshot AI credits partnership reads like a funded team testing a wedge. Star velocity over the next 30 days is the cheap tell.
- What does Goodnotes' Sept 2025 "generate visuals... directly in your canvas" release actually ship? The direct fetch failed. If Goodnotes already places open-ended generated content on the page, the incumbent threat is materially larger than this brief scores it.
- Is there an App Store category gap that can be measured? No pen-first agentic canvas appears to exist on iPad. Confirming that against App Store search results would turn the strongest claim in this brief from inference into evidence.
- What is the actual willingness-to-pay for a metered-token creative tool among non-developers? The Goodnotes AI Pass price is an anchor, not a demand curve, and the whole "option 4" plan rests on it.
- Did Muse/Allume fail for reasons that generalize? The best-designed pen-forward thinking canvas of the decade unwound its venture structure. That post-mortem is directly load-bearing for ink and is a single vault note away.
Related
- [[2026-07-21-ink-build-spec]]
- [[IMPLEMENTATION-NOTES-2026-07-21-ink-v0]]
- [[2026-05-04-karlmehta-llm-commoditization-intelligence-rails]]
- [[2026-05-13-fde-asymmetric-edge-rdco-positioning]]
- [[project_l5_north_star_strategic_direction]]
Sources
Vault
- [[2026-07-21-ink-build-spec]] —
~/rdco-vault/01-projects/ink/2026-07-21-ink-build-spec.md - [[IMPLEMENTATION-NOTES-2026-07-21-ink-v0]] —
~/rdco-vault/01-projects/ink/IMPLEMENTATION-NOTES-2026-07-21-ink-v0.md - [[2026-05-04-karlmehta-llm-commoditization-intelligence-rails]]
- [[2026-05-13-fde-asymmetric-edge-rdco-positioning]]
Web — fetched vendor pages (prices verified)
- https://www.goodnotes.com/pricing
- https://remarkable.com/shop/connect/pricing · https://remarkable.com/products/remarkable-paper/pro · https://remarkable.com/products/remarkable-paper/pro-move
- https://www.notability.com/pricing
- https://miro.com/pricing/
- https://www.figma.com/pricing/
- https://plus.excalidraw.com/pricing
- https://tldraw.dev/pricing · https://tldraw.dev/docs/ai
- https://allume.com/ (301 target of museapp.com)
- https://api.github.com/repos/penecho/penecho (repo metadata: created 2026-07-14, 1,701 stars, AGPL-3.0)
- https://raw.githubusercontent.com/penecho/penecho/main/README.md
Web — launch/feature sources
- https://www.goodnotes.com/blog/introducing-goodnotes-6 · https://support.goodnotes.com/hc/en-us/articles/14500614035471-Spellcheck-your-handwriting
- https://www.apple.com/newsroom/2024/06/ipados-18-introduces-powerful-intelligence-features-and-apps-for-apple-pencil/
- https://www.figma.com/blog/introducing-figma-make/ · https://www.figma.com/blog/welcome-weavy-to-figma/
- https://community.miro.com/product-news-31/what-s-next-with-miro-introducing-the-intelligent-canvas-17071
- https://tldraw.dev/blog/make-real-the-story-so-far
- https://adamwiggins.com/muse-retrospective/
- https://wiki.heptabase.com/roadmap/
- https://techcommunity.microsoft.com/blog/microsoft365insiderblog/onenote-copilot-now-supports-inked-notes/4226693
- https://github.com/MaximeRivest/riddle (prior art credited by ink)
Flagged — could not verify, do not cite as fact
- BusinessWire, "Goodnotes Becomes First to Bridge Digital Ink and Generative AI" (2025-09-23) — fetch timed out; canvas-generation scope unverified
- Microsoft Research ImaginationVellum (UIST '25) — microsoft.com and dl.acm.org returned 403; mechanics from abstract only
- Concepts app pricing (~$5/mo, $30/yr) — vendor pricing page 404'd, search-sourced only
- MyScript Notes pricing — no dollar figure on vendor page
- Napkin AI, Whimsical, Heptabase, Kosmik, Boox, Supernote, Kindle Scribe prices — search-sourced, not fetched
- Spine AI (getspine.ai) revenue and YC batch — conflicting third-party attributions