Skip to content

For Canadian teams documenting SR&ED work

Build a stronger SR&ED record as the work happens.
Start with dated evidence, not year-end memory.

T661.ai organizes GitHub history and facts supplied by your team into monthly technical records, annual narrative drafts, checked expenditure worksheets, and a package your accountant can review.

Create my free previewSee a sampleOne free completed-month preview · no card required · 14-day trial starts only when the preview contains evidence
Read-only GitHub AppNo source files or code diffs storedYour team reviews every final recordCancel any time

Design-partner release: finalization currently supports straightforward British Columbia claims from Canadian-controlled private corporations using the salary-based proxy method. Other claims can still build and export drafts for an accountant. Check the supported boundary.

Why monthly SR&ED evidence matters

01

A successful claim can extend your runway

Good technical records help protect the value of your R&D work. The software fee should not grow just because the resulting credit does.

02

Missing records make review harder

CRA says the best supporting evidence is dated and specific to the work. Rebuilding the story at year end is harder than keeping clear records as the work happens.

03

Useful evidence may already be in GitHub

Commits can show experiments, benchmarks, reversals, and failed approaches. T661.ai organizes those signals by month and clearly shows where your team must add context.

How it works

  1. 1

    Choose the GitHub repositories with R&D work

    Install the read-only GitHub App and select only the repositories you want T661.ai to use. A repository is where code is stored; a technical project is the specific technical challenge your company investigated. One project can use evidence from several repositories or selected monorepo paths.

  2. 2

    T661.ai organizes the dated evidence

    Every branch is scanned. Each commit is assigned to the month it was originally authored, duplicates are removed, and experimental work, benchmarks, reversals, and abandoned approaches are separated from routine maintenance.

  3. 3

    AI drafts; your team verifies

    Claude groups the month's experimental commits into testable ideas and summarizes the method, result, and learning. Each statement links to commit IDs (SHAs). T661.ai checks those links, and your team corrects the facts, resolves every ⚠ VERIFY marker, enters hours from company records, and confirms the final log.

  4. 4

    Give your accountant an organized package

    Download a locked ZIP snapshot with the T661 technical narrative drafts, expenditure worksheet, monthly logs, underlying evidence, supporting documents, revisions, and a file-integrity list. It is a working-paper package—not an EFILE submission or tax opinion.

acme-robotics__2026-03.md
## 6. Work Performed — Iterations (March 2026)

| Hypothesis / Approach | Method | Results & Metrics |
|---|---|---|
| **H1** — Streaming TTS chunking at clause
boundaries will cut perceived latency below
800ms without prosody artifacts | Implemented
clause-boundary splitter (`a3f81c2e`,
`9d04bb71`); benchmarked against sentence-
level baseline (`c2e94d10`) | p95 first-audio
1.9s → 740ms. Prosody regression at clause
joins — reverted joiner (`f81d3a55`) |

Real structure, fictional content. Each technical statement links to a commit ID so a reviewer can trace the evidence. That traceability supports review; it does not determine eligibility.

More than a log generator

One workspace from the first commit to accountant review.

Monthly evidence is the foundation. T661.ai combines it with facts from the company making the claim, structured review steps, and a locked package of working papers.

Inspect the public sample →

Monthly evidence

Dated commits, failed approaches, benchmarks, and links back to each cited commit.

Hours from company records

Your team enters hours and names the source before seeing the GitHub activity comparison.

Annual technical narratives

Drafts for T661 Part 2, with word limits, continuity between years, and required checks.

Separate review and confirmation

Team comments, external reviewer approval, and final confirmation by the claiming company.

Expenditure worksheet

Supported calculations using each person's compensation, project allocation, available time, and the applicable rates.

Locked accountant handoff

A fixed ZIP snapshot with evidence, revisions, supporting documents, and hashes that detect changed files.

Organized around CRA's two requirements

CRA's current guidance asks why the technical work was needed and how the investigation was carried out. Your company explains the technical knowledge gap. Dated repository evidence can support the record of planned experiments or analysis.

Why

What could the team not determine in advance?

Name the technical limits and explain why standard approaches did not resolve them. The project does not need to succeed to be documented.

How

What testable idea did you investigate?

Record a specific prediction, not a list of features.

How

How did you test it?

Record the method, result, and learning, with links to the relevant commit IDs.

How

What did the result teach you?

State what became known, what became possible, and which approaches were abandoned.

How

What records were kept?

Use dated commits and other company records created while the work was happening.

The Income Tax Act lists the excluded activities at 248(1)(e)–(k): market research or sales promotion, quality control or routine testing, social sciences and humanities research, prospecting for minerals or petroleum, commercial production, style changes, and routine data collection. “Routine engineering” is not on it — engineering, design, and computer programming are named support-work categories. Every exported claim package includes the complete list for professional review.

Current product fit

Check whether the current version fits your claim.

T661.ai is in a design-partner release. Most teams can organize and export their evidence, but only the straightforward claim type below can be finalized inside the product today.

Supported finalization path today

A British Columbia Canadian-controlled private corporation (CCPC) using the proxy method, where overhead is calculated from eligible salary. The claim must not involve associated corporations, materials, capital property, late-paid amounts, work performed outside Canada, or work performed for another party.

Other claim types: export for accountant review

Teams in other provinces, teams using the traditional overhead method, and claims with any complexity listed above can still build and export their evidence. T661.ai keeps the annual claim in draft so an accountant can complete the unsupported tax treatment.

“Ready for accountant review” means the files are organized and the product's required checks have passed. It does not promise that an accountant or CRA will accept the claim without changes.

Compare software pricing with percentage fees

T661.ai charges for the software, not a share of the credit. Enter an expected credit and any percentage-fee quote you received to compare the two costs.

Use your own numbers

Compare a percentage fee with 12 paid months of T661.ai

Enter the credit amount you expect and the percentage you were quoted. This compares fees only; it does not estimate your claim or decide eligibility.

The 15% starting point is illustrative and user-editable. Published Canadian provider pages describe about 15% as an average and a wider 10–30% typical range. Replace it with your actual written quote.

T661.ai plan

Percentage fee

$18,000

15% of the entered credit

T661.ai for 12 paid months

$588

at today's Solo monthly price, before tax

Estimated fee savings

$17,412

96.7% lower than the entered percentage fee

At 15%, the two prices break even at an estimated credit of $3,920.

Fee-only comparison in CAD, before tax. It assumes 12 paid months at the current price and excludes internal time, tax-return preparation, optional professional review, and audit support. Software and professional services have different scope. This is not a credit estimate, savings guarantee, or eligibility conclusion.

Pricing

Flat monthly software pricing in Canadian dollars. T661.ai never takes a percentage of your claim. Both plans include the same workflow; the only difference is how many active repositories you can connect.

Start with one free preview for one project and one completed month, using up to three repositories. Choose a plan only after the preview finds evidence and creates a log.

Solo

$49/month

  • Up to 3 repositories
  • Monthly logs plus any earlier completed month
  • AI-assisted drafts with required human review
  • Accountant ZIP with revisions and file-integrity checks
  • Unlimited members in each company workspace

Team

$99/month

  • Up to 10 repositories
  • Monthly logs plus any earlier completed month
  • AI-assisted drafts with required human review
  • Accountant ZIP with revisions and file-integrity checks
  • Unlimited members in each company workspace
Create the free preview first

FAQ

What does CRA actually require?

CRA looks at two things: why the work was needed and how the investigation was carried out. CRA also says the strongest supporting records are dated and specific to the work. T661.ai preserves the original dates from Git history and clearly labels any later AI-written draft. GitHub evidence can support your account of the work, but it does not decide eligibility. Your company still provides the technical context, reviews every statement, and confirms hours from its own records.

Does T661.ai read my source code?

The GitHub App needs read access because GitHub uses that permission for commit information. T661.ai stores evidence about the work—such as commit IDs, messages, authors, dates, branches, file paths, change counts, and relevant pull-request text. It does not store source files or diffs. If GitHub includes a code fragment in an API response, T661.ai discards it immediately and never sends it to Claude.

Is the AI going to invent results?

Claude creates a draft; it does not decide what is true or whether work is eligible. T661.ai checks every cited commit ID and adds a ⚠ VERIFY marker when the evidence does not support a statement. You must correct or remove every marker, enter hours from company records, and make the final confirmation yourself.

Is this tax advice? Will this guarantee my claim?

No and no. T661.ai produces technical working papers from your repository history. Eligibility determinations, expenditures, and filing remain with you and your SR&ED advisor or accountant. The product preserves dated evidence, exposes unsupported statements, and blocks incomplete packages inside its supported scope.

What if my repos aren't on GitHub?

GitHub is the only repository provider supported today, through a read-only GitHub App. Support for self-hosted Git, GitLab, and Bitbucket is planned. Tell us what you use so we can prioritize it.

Can I backfill months from earlier in my fiscal year?

Yes. You can generate a log for any completed month covered by your Git history. The original commit dates stay intact, and T661.ai separately records when the later summary was generated and reviewed. After setup, T661.ai schedules a new draft when each month ends and alerts you if it cannot finish.

What do I hand to my accountant?

A locked ZIP snapshot containing the annual technical narrative drafts, expenditure worksheet, monthly activity logs, underlying evidence, supporting documents, revision history, and a file-integrity list. It is an organized set of working papers—not a filed tax return or a tax opinion.

Why keep using it after the first draft?

Monthly records are easier to review than a year-end reconstruction. New work is captured while context is fresh, failed approaches remain visible, deadlines stay in view, and every revision and supporting document stays connected to the right claim year.

T661.ai — Evidence-backed SR&ED working papers