Copilot usage metrics: agent activity is not your bill
GitHub adds client and custom-tool breakdowns to Copilot usage reports. Check attribution, permissions and missing telemetry before interpreting AI agent activity.
Fewertools Watch
What the providers' public pages say now, what changed from our dated notes, and what still needs checking before you choose.
The source notes below check public documentation. The workflow records describe the exact synthetic tasks executed and their limits. Older approach reviews keep their original dates and remain historical until we review them again.
Latest tool and pricing checks · RSS feed · Route guides · Get updates
Published research
Saved research index from 26 October 2026. Check the latest published updates. Individual source and publication dates are unchanged.
GitHub adds client and custom-tool breakdowns to Copilot usage reports. Check attribution, permissions and missing telemetry before interpreting AI agent activity.
beehiiv’s October roundup adds editor and sending controls. Check list settings, scoped API keys and conflicting automation eligibility before changing a workflow.
Nine scoped re-reads cover beehiiv, Framer, Gumroad, CapCut, Linear, Notion and Webflow. Billing bases and untested workflow boundaries remain explicit.
A dated checklist for software allowances, trial expiry and infrastructure costs, using official Fly.io, Auphonic, Artillery and Authentik sources.
Copysmith platform scope, Eppo/Datadog and Auth.js identity, plus Authentik, Auphonic and Artillery free-access boundaries. Paid prices and editorial scores are unchanged.
Scoped pricing tables, documented free-plan limits and 9 newly sourced catalogue marks. Student searches connect to documented editing capabilities; qualified brand labels identify providers and legacy projects.
Three scoped pricing tables, access/identity corrections and 27 newly sourced catalogue marks. Qualified brand labels distinguish providers and renamed products.
Four scoped pricing tables, access/identity notes and 42 newly sourced catalogue marks. Qualified brand labels distinguish providers and renamed products.
19 refreshed pricing tables, scoped access notes and 406 newly sourced catalogue marks. Provider/family logos are labelled and unresolved sources retain initials.
Selected plan limits, billing displays and current pricing snapshots, with separate official icon provenance and unresolved source scopes.
One IES report through Zotero, a Google Docs outline and three Anki cards. Includes the metadata correction, native exports and explicit limits.
A shared canvas is a team view. Paid-plan eligibility, privacy settings and repository access still apply.
The new presentation workflow depends on Space access. Exports and Windows local editing have specific limits.
What evidence would let you replace an inherited meeting-summary workflow without changing its decisions, owners or conditions?
Three models have a retirement date. One has an exception for annual individual subscribers.
A single response can involve several models. The preview’s billing and workspace limits matter as much as the model picker.
A documented connection selector and checklist for runtime, pooling and access boundaries.
A local calculator for overlapping subscriptions, one-off setup and the planning-period total.
The same model has different subscription and usage-billing routes. Here is what the 13 October announcements actually establish.
Enterprise settings can attribute activity to an account without enabling prompt or tool-content capture.
Source checks and partial browser observations for Pocket Invoice, IcyZip and Aura by Deslove. Read free-access boundaries, account gates and unverified workflows.
Check the IDE version first, then separate model defaults from the tools that start automatically.
The beta suggests a follow-up after a response. Accepting the text leaves the decision to send with you.
Separate Claude Code instructions, tool permissions and shell sandboxing. Inspect a synthetic settings file with a local review script and a practical checklist.
Run a reproducible Git worktree example with separate working files and a shared starting commit. Learn what worktrees isolate and what they still share.
Inspect a newsletter CSV before migration with a synthetic 16-row fixture, duplicate and status checks, saved results and a read-only Python validator.
Separate raw contacts, active labels, confirmed opt-ins and billable subscribers. Use an executed synthetic count example and a provider-scope worksheet.
Calculate transcription WER from substitutions, deletions and insertions. Reproduce seven local text examples and see why one score cannot establish accuracy.
Inventory PCM WAV duration locally and build a scoped transcription budget. Reproduce silent fixtures while keeping vendor rates and unobserved costs unknown.
Explore fourteen dated beehiiv, Kit and Buttondown cost scenarios. Separate monthly quotes, annual arithmetic and promotions, with unknown renewal totals preserved.
Cursor highlights iOS control of local agents. Its current docs add important cloud-context and team-access requirements.
Rollouts follows deployed changes using telemetry; Security Review examines pull requests. Neither is evidence that a deployment is healthy without its signals.
Local sandboxing is generally available across supported Copilot surfaces. GitHub’s docs say it is off by default and settings are separate in CLI and app.
The model picker can discover a running Ollama instance. It does not install a model, switch automatically or enable offline mode.
Organizations can choose a billing source for eligible code reviews and control externally licensed requests. Access and runner costs remain separate.
ElevenLabs’ v4 launch includes a low-latency Turbo variant. Its request-to-speech benchmark removes network latency; it is not a Fewertools measurement.
Preview gains a Gmail clipping estimate. Nearby fixes cover mobile alt-text checks and unsubscribe links; verify your own reader journey before relying on them.
Source-checked beehiiv, Kit and Buttondown quotes at 1,000 and 10,000 readers, feature gates and export limits. Fourteen local cost scenarios are reproduced with synthetic fixtures. Provider account sends, imports, exports and payment journeys remain pending.
New synthetic workflow records cover Figma, Google Docs, Google Sheets, GitHub and Tally. An agent executed the workflows and published reproducible fixtures and scoped results. Tally covers a simpler preview form only; its planned conditional form remains partial. These records do not change editorial scores or reopen historical route reviews.
Public-document observations
Nine scoped documentation re-reads of the 9 October notes. This is not a refresh of all 45 historical route claims or a workflow test. Official sources read 26 October 2026. Account access, invoices and complete workflows were not tested. Historical route eligibility and September expiry boundaries are unchanged. The next-check date applies only to these notes.
NarrowedbeehiivDocument observed 26 October 2026
Use the current plan comparison. Automation eligibility differs between the 23 October roundup and pricing copy; confirm the required feature before paying.
ConfirmedFramerDocument observed 26 October 2026
A one-time allowance is not a monthly budget or a guaranteed number of finished sites.
ConfirmedGumroadDocument observed 26 October 2026
These headline platform fees do not establish every seller’s total payment, refund or payout cost.
ConfirmedCapCutDocument observed 26 October 2026
Check the timeline and export settings. This does not establish free access to all premium assets, formats or AI features.
ConfirmedLinearDocument observed 26 October 2026
Selected annual quotes only. Keep AI-credit costs, seats and attachment limits separate.
ConfirmedNotionDocument observed 26 October 2026
The monthly/annual profile observations retain their own scope. AI credits and add-ons are separate; no account use was tested.
ConfirmedbeehiivDocument observed 26 October 2026
These are annual-billing equivalents, not monthly-billing quotes. No invoice, subscriber growth or delivery was tested.
ConfirmedFramerDocument observed 26 October 2026
Editor seats, AI credits and usage can add cost. These figures do not establish a working website output.
ConfirmedWebflowDocument observed 26 October 2026
Workspace seats, ecommerce and cloud usage are separate. No publishing workflow or complete project cost was tested.
These records preserve their original dates. Later profile checks cover their stated facts; they do not renew every earlier claim.
Public-document observations
Nine public-document observations: four bounded provider-claim notes plus five corrections from the twenty-tool pricing recheck. This is not a refresh of all 45 historical route claims, the 12 reference breakdowns or any workflow permission. Official public pages were read on 9 October 2026. Notion USD annual/monthly billing was also inspected in the vendor UI. No provider account, purchase, generation, newsletter delivery or provider API was used. Account-specific access, output quality and total cost remain untested. Historical route eligibility and September expiry boundaries are unchanged. The next-check date applies only to these notes. All twenty bounded pricing observations, including their currency and billing scope, are published in the directory pricing receipt.
Historical snapshot. These nine notes were revisited on 26 October 2026. Read the newer scope and source observations above.
ChangedbeehiivDocument observed 9 October 2026
Choose against the current plan table, not the September names or credit numbers. Subscriber count and billing period change the displayed price; a paid trial is not a continuing free allowance. No account allowance or newsletter send was tested.
ChangedLinearDocument observed 9 October 2026
Budget per paid user and keep the annual billing condition. Monthly billing, AI-credit cost and hands-on issue tracking were not checked in this observation.
ChangedNotionDocument observed 9 October 2026
Compare the same currency and billing period. Custom Agent credits and other add-ons remain separate, and product use was not tested.
ChangedbeehiivDocument observed 9 October 2026
Subscriber count changes paid pricing. This verifies published plan prices, not subscriber growth, delivered emails or account access.
ChangedFramerDocument observed 9 October 2026
Separate the site subscription from editor seats, AI credits and usage. These prices do not establish a working website output or total project cost.
ChangedWebflowDocument observed 9 October 2026
A Site subscription differs from Workspace seats, ecommerce, cloud usage and platform contracts. The review now states the billing basis and uses the current Site names. Publishing was not tested.
NarrowedFramerDocument observed 9 October 2026
The Free allowance is not monthly. Additional credits have purchase and expiry conditions, and prompt cost varies, so a credit balance cannot promise a number of completed sites. This was a documentation check, not an Agent or website test.
NarrowedGumroadDocument observed 9 October 2026
Use the published platform fee as one input, not a guaranteed net payout. The detailed fee-help page returned no readable body in this check. Seller approval, local payout eligibility and an actual purchase/download path remain untested.
ConfirmedCapCutDocument observed 9 October 2026
The current public help page still supports this bounded guidance. It does not establish free access to every premium asset, export format or AI feature. No project was opened and no video was exported.
Historical route-evidence recheck
The 13 route claims due for recheck by 10 September 2026. The other 32 claims were not due and were not reassessed. Each provider's current public page was opened directly, starting from the recorded source URL. No account, generation, paid test or provider API was used.
Historical evidence. This review is past its recheck date. These original claims have not been renewed by the documentation notes above. You can build a working brief while an approach needs rechecking.
ChangedFramerChecked 7 September 2026
If you planned to build a Framer site over several months on the Free plan, the credits do not come back. Only paid plans renew, on their billing date. The help page was updated on 1 September 2026 and the pricing page corroborates the 500 introductory credits.
NarrowedCapCutChecked 7 September 2026
The no-automatic-watermark part stands. Paid-export conditions were not substantiated by the watermark article, so they now live with the credits evidence instead. Check the whole timeline for template branding before you export.
NarrowedCapCutChecked 7 September 2026
If you edit on a phone, the credit rules on this page do not apply to you yet, and the page does not say what does. Plan credit-consuming AI steps on Web or Desktop.
NarrowedCapCutChecked 7 September 2026
A Commercial Sound is not cleared for YouTube or Instagram unless you license it separately. The agreement's introduction sends US users to a different document, which this recheck did not verify.
NarrowedbeehiivChecked 7 September 2026
Do not assume the free Launch plan lets you draft an issue with beehiiv AI. The numbers are on the page; the access is not, and this recheck could not resolve the contradiction without an account.
NarrowedCanvaChecked 7 September 2026
The allowance is one shared pool, so a complex task can use several credits, and some AI design tools do not draw from it at all. Budget on the pool, not on 200 separate tries.
NarrowedOpenAIChecked 7 September 2026
Turning off training is not an absolute promise. A thumbs-up or thumbs-down on a conversation can permit training on that conversation. Use a Temporary Chat for anything you would not want trained on.
NarrowedOpenAIChecked 7 September 2026
The rule is scoped to uploading images or video of a person. It still means you need consent before reproducing someone's likeness that way. Do not read it as a statement about every text-only request.
ConfirmedCanvaChecked 7 September 2026
The page now states an effective date of 26 June 2026. The substance is the same: outputs can overlap with other people's, are not verified for accuracy, and your inputs can be shared with Canva's technology partners.
ConfirmedCanvaChecked 7 September 2026
A Pro photo or element is cleared for the design you export it in, not as a file you can resell or reuse. Some Free Content can be downloaded standalone, which is a separate section of the agreement.
ConfirmedFramerChecked 7 September 2026
The help article, updated 7 August 2026, recommends working one section at a time. This confirms documented capability only; no generated site was tested for quality.
ConfirmedOpenAIChecked 7 September 2026
Documented capability, not a quality claim. The source page changed its title, which is why the recheck is recorded even though the substance is unchanged.
ConfirmedWixChecked 7 September 2026
Nothing to change in the route. No account was created for this recheck, so plan-specific behaviour inside the builder remains untested.
The Drop archive
Watch and The Drop now share this research log. These earlier posts keep their original publication dates and have not been rechecked as part of the current pricing review.
The Drop
Watch is the public research archive. The Drop sends new notes from this archive by email. Subscribe below, or use the RSS feed.
The Drop
Subscribe to The Drop from Fewertools. You will receive a confirmation email first. This is separate from the Editions collection; unsubscribe in any email.
Contact FewertoolsPublished research
Google adds in-chat file creation and editing. PDFs work on web and mobile; the other listed formats are web-only.
Google announced Carbon on 28 October, starting with selected cyber defenders and trusted partners. Broad product availability is still planned.
GitHub’s Universe update adds Projects in public preview, including work across repositories. Desktop computer use has separate controls.
OpenAI says it will honour a shared robots.txt signal in the coming months. Existing training and search crawler controls remain separate.
OpenAI’s life-sciences model is rolling out to approved Research organisations in Work, Codex and separately approved API access, not the Chat tab.
A stable AI Gateway model ID can point to changing models and settings. Shared configuration makes rollout and provider checks important.
Updated scope: Haiku 5.5 replaces Haiku 4.5 in consumer chat, Cowork and Design selectors. Code, Team, Enterprise and API access are unaffected.
Updated scope: OpenAI’s 29 October response-style change affects Free, Go and signed-out Chat, plus the limited-message fallback on Plus and Pro.
ChatGPT is rolling out two-account switching on iOS and Android. Chats, memory and subscriptions remain separate; availability can vary by account.
Resend adds shared inboxes for teams and agents. Three inboxes are included, but their incoming and outgoing email uses the existing transactional quota.
Starting allowance, full-book purchase boundary and separate monthly/annual prices. Three official icons added.
Eligible enterprises can adopt Copilot access without a separate seat fee for licensed users. Metered usage starts immediately; self-serve plans are unchanged.
Free-plan limits, trial boundaries and separate UK monthly/annual Captivate quotes. Unknown paid amounts remain unknown.
Pinned local code, synthetic rows and observed RLS outcomes. Also corrected Glitch’s retired hosting status and checked Buffer, Capacities and Gleap plan boundaries.
Updated Claude and Copilot billing scopes, a blank cost worksheet and a Jira migration checklist. The Jira guides disagree on sub-team eligibility; the limitation is explicit.
Claude gains an Anthropic-managed deployment option. OpenAI models enter a separate private preview. Processor, feature access and billing need separate checks.
Google announced a business hub, reusable brand styles and connected apps. A global rollout announcement does not confirm every account or integration is ready.
Boot.dev and Brand24 billing, Brevo daily sending, Brave Leo access, Boomi trial limits and Bruno editions. Fourteen official logos added.
The new organisation-owned identity is in public preview. Its read-only permission can consume AI credits, and per-app budgets are not available.
The standard moves to a neutral foundation. The file format is portable; runtime requirements, permissions and successful execution still need checking.
Bigin GBP billing, Bitwarden annual USD prices, Bitbucket shared free limits, Blender and BookStack software costs, and Blush export access. Ten product-specific official marks added.
A five-field permission worksheet for a coding agent or newsletter integration, grounded in current Cursor and Buttondown documentation.
Code-defined agents have distinct deployment and ownership roles. The current SDK package instructions differ from the launch post.
The provider says email-creation requests that would send now require sending access. Draft creation remains separate.
Three dated pricing checks, including a conflicting Team billing unit and explicit free-versus-trial limits.
Newly recorded USD annual quotes, editor billing, project allowances and trial limits. Monthly quotes and hands-on workflows remain unverified.
GitHub adds client and custom-tool breakdowns to Copilot usage reports. Check attribution, permissions and missing telemetry before interpreting AI agent activity.
beehiiv’s October roundup adds editor and sending controls. Check list settings, scoped API keys and conflicting automation eligibility before changing a workflow.
Nine scoped re-reads cover beehiiv, Framer, Gumroad, CapCut, Linear, Notion and Webflow. Billing bases and untested workflow boundaries remain explicit.
A dated checklist for software allowances, trial expiry and infrastructure costs, using official Fly.io, Auphonic, Artillery and Authentik sources.
Copysmith platform scope, Eppo/Datadog and Auth.js identity, plus Authentik, Auphonic and Artillery free-access boundaries. Paid prices and editorial scores are unchanged.
Scoped pricing tables, documented free-plan limits and 9 newly sourced catalogue marks. Student searches connect to documented editing capabilities; qualified brand labels identify providers and legacy projects.
Three scoped pricing tables, access/identity corrections and 27 newly sourced catalogue marks. Qualified brand labels distinguish providers and renamed products.
Four scoped pricing tables, access/identity notes and 42 newly sourced catalogue marks. Qualified brand labels distinguish providers and renamed products.
19 refreshed pricing tables, scoped access notes and 406 newly sourced catalogue marks. Provider/family logos are labelled and unresolved sources retain initials.
Selected plan limits, billing displays and current pricing snapshots, with separate official icon provenance and unresolved source scopes.
One IES report through Zotero, a Google Docs outline and three Anki cards. Includes the metadata correction, native exports and explicit limits.
A shared canvas is a team view. Paid-plan eligibility, privacy settings and repository access still apply.
The new presentation workflow depends on Space access. Exports and Windows local editing have specific limits.
What evidence would let you replace an inherited meeting-summary workflow without changing its decisions, owners or conditions?
Three models have a retirement date. One has an exception for annual individual subscribers.
A single response can involve several models. The preview’s billing and workspace limits matter as much as the model picker.
A documented connection selector and checklist for runtime, pooling and access boundaries.
A local calculator for overlapping subscriptions, one-off setup and the planning-period total.
The same model has different subscription and usage-billing routes. Here is what the 13 October announcements actually establish.
Enterprise settings can attribute activity to an account without enabling prompt or tool-content capture.
Source checks and partial browser observations for Pocket Invoice, IcyZip and Aura by Deslove. Read free-access boundaries, account gates and unverified workflows.
Check the IDE version first, then separate model defaults from the tools that start automatically.
The beta suggests a follow-up after a response. Accepting the text leaves the decision to send with you.
Separate Claude Code instructions, tool permissions and shell sandboxing. Inspect a synthetic settings file with a local review script and a practical checklist.