Why this exists.
Infrastructure work runs through tools that assume you can see the screen. Terraform, OpenTofu, Ansible, the AWS CLI and gcloud print long, dense output, use colour to mark what is dangerous, and draw tables that a screen reader reads cell by cell. A deletion in a production plan sounds the same as a tag change. A blind or low-vision engineer can do the work, but has to dig for the one line that matters.
These tools start from the other end. They are designed for engineers who use screen readers, braille displays, magnification or large print, and rcdo is being designed by a visually impaired operator. The most important fact comes first, as labelled text, at a width you choose. That makes the output faster for everyone else to read too.
Design principles.
- Plain text first. Colour is optional and never the only signal. A class label such as
[vm]is always in the text, even when colour is on. - What matters first. Findings lead with severity, resource, action, environment and reason, rather than leaving them at the end of a long block.
- Layouts for your output device. rcdo has plain, speech and braille layouts, with braille at 40 columns by default. Review commands take
--widthdown to 40 columns, and nothing depends on a visual table. - Navigable, resumable reading. Long reports can be walked one finding at a time in severity order, with bookmarks, notes and a saved position. If the source changes underneath you, the session says it is stale rather than letting you continue on old evidence.
- Incomplete is never clean. Missing, stale or invalid evidence makes a review INCOMPLETE, with its own exit code. A partial check is never reported as a pass.
- Structured output underneath. Alongside text, the tools emit versioned JSON with a named schema, so the same evidence can feed a screen reader, a braille display, CI annotations or another tool.
- Read-only by default. Anything that changes a file or a system needs an explicit flag, and edits show a preview first.
- Offline and simple. The tools work with ordinary files and pipes, need no daemon, and work without AI. rcdo’s
ai-assistcommand can send redacted input to an AI provider you configure, but only when you opt in.
The tools.
rcdo
Reads Terraform, OpenTofu, Ansible and cloud output and reports what matters as labelled findings, with speech and braille layouts, resumable review sessions and shift briefs. Go, GPL-3.0-or-later.
Collect evidenceMissing Utils
48 small command-line tools that collect host, network, process, cloud and IaC evidence as stable JSON, and say plainly what they could not see. Go, MIT.
Read requestsprtool
Shows GitHub pull requests and GitLab merge requests as labelled fields and a diff whose + and - markers stay in the text. Read-only. Go; no license file yet.
Missing Utils also publishes its command list as a linear text file, wrapped for speech and magnification, alongside the JSON inventory.
How they work together.
Missing Utils does bounded collection and normalisation. rcdo does the nonvisual side: navigation, previews, interpretation, review state, runbook progress and handoffs. Four rcdo commands read Missing Utils evidence directly: context reads a contextsnap snapshot, watch reads an eventwhy artifact, receipt-review reads runreceipt evidence, and jsonprobe-check reads jsonprobe readiness results. Each tool also works on its own. The two projects share one accessibility roadmap.
Try it.
Review a saved OpenTofu plan as a 40-column brief before applying it. Nothing is applied:
tofu plan -out=review.tfplan
rcdo tofu-check --plan review.tfplan --environment production --format json > tofu-review.json
rcdo review-brief --input tofu-review.json --width 40Check service readiness with Missing Utils, then have rcdo decide whether the required checks passed and are fresh:
jsonprobe --input probes.json > readiness-source.json
rcdo jsonprobe-check --input readiness-source.json --require api-health \
--environment staging --max-age 15m --format jsonjsonprobe exits 1 when a probe fails but still writes its results, so keep the file either way. jsonprobe-check exits 0 when required checks pass, 20 when one fails, and 30 when evidence is missing or stale. The readiness workflow shows how to handle each exit code safely. Install steps are in the rcdo and Missing Utils guides.
ADA, Section 508 and WCAG.
There is no ADA certification for software, and a tool cannot make an organisation compliant by itself. The Americans with Disabilities Act asks employers to provide reasonable accommodation. Section 508 and WCAG set accessibility requirements for technology and web content. Tools that a blind engineer can use independently make that accommodation practical in operations work, and these are built for that purpose.
WCAG is written for web content rather than command-line output, but its principles carry over, and they guided the design: information is perceivable without colour or layout, operable from the keyboard and a pipe, understandable through labelled findings, and robust because the same evidence is available as structured data for any assistive technology.
No compliance claim. We do not describe these tools as ADA or Section 508 compliant. They are designed for accessibility, and we will publish real-world assistive-technology results as they are measured.
What has been verified.
| Area | Status |
|---|---|
| Output width and control characters | Checked automatically, including plain-text output at 72 columns in the readiness demonstration. |
| Builds and tests | Both toolkits build and pass their test suites. |
| Screen-reader, braille-display and magnification use | Not yet measured. Automated checks do not establish it. rcdo pilot records an operator’s acceptance pass. |
| Authenticated cloud and workplace workflows | Pending. No live employer credentials were used during development. |
| Releases | No tagged release yet. rcdo is a beta core (1.3.0-beta.1); Missing Utils is unreleased. Install from source. |
Both repositories keep a current record of what is done and what is still open: the rcdo delivery status and the shared accessibility roadmap.
Help us test.
If you use a screen reader, a braille display, magnification or large print for infrastructure work, your feedback is what turns “designed for” into “verified with”. Tell us what you use and where these tools get in your way: open an issue on rcdo or Missing Utils, or email us.