AmICited CLI

Inspect, clean, verify, and safely rewrite AI-assisted text before you publish it.

Explore AmICited · AI visibility features · AI search insights

AmICited CLI is the open-source, text-only watermark toolkit from AmICited, the AI-search visibility platform by FlowHunt. It helps content teams and AI agents find deterministic artifacts, produce reviewable rewrites, preserve important spans, and record exactly what changed.

Use it to improve the technical hygiene and human review of AI-assisted content before it enters your publishing workflow. Removing hidden characters or rewriting text can reduce specific detectable artifacts, but it does not guarantee evasion of Google or another proprietary system, human authorship, or higher search rankings. Strong search performance still depends on useful, original, accurate, and authoritative content.

Want to know whether AI search cites your brand? Run your visibility check with AmICited →

What it does

• Inspects locally by default for hidden Unicode, bidi controls, Unicode tags, exotic spaces, confusables, normalization differences, and suspicious whitespace.
• Removes deterministic artifacts safely while preserving the original, line endings, Markdown structure, and a complete change record.
• Creates protected rewrites through an API model, Codex CLI, or Claude CLI while protecting citations, URLs, quotations, numbers, frontmatter, and code.
• Verifies before and after using detector-specific statuses instead of a manufactured global confidence score.
• Works for people and agents through versioned JSON, a Python SDK, and the bundled amicited-watermarks skill.

Install

Install the amicited CLI in an isolated environment with uv:

uv tool install amicited
amicited watermark capabilities

Run it once without a persistent installation:

uvx amicited watermark capabilities

Upgrade later with uv tool upgrade amicited. AmICited requires Python 3.11 or newer; uv can provision a compatible interpreter automatically.

Quick start

# Inspect without changing the source
amicited watermark inspect article.md > article-inspection.json

# Produce article_dewatermarked.md and a structured report
amicited watermark rewrite article.md > article-rewrite-report.json

# Use an existing authenticated Codex session for semantic rewriting
amicited watermark rewrite article.md --provider codex > article-rewrite-report.json

# Verify the transformed file against supported deterministic signals
amicited watermark verify article_dewatermarked.md

Agent skill

Install the bundled skill globally without cloning this repository:

amicited watermark skills

Choose Codex or Claude when prompted, or select one directly:

amicited watermark skills --provider codex
amicited watermark skills --provider claude

Codex installs globally at ~/.agents/skills/amicited-watermarks; Claude Code installs globally at ~/.claude/skills/amicited-watermarks. Running the command again is safe when the bundled and installed copies match. If an existing copy differs, AmICited refuses to replace it. Use --force to preserve the old copy as a timestamped backup and install the bundled version.

The amicited-watermarks skill teaches agents to use file-first input, inspect before rewriting, preserve the source, request confirmation before external processing, and interpret verification without overstating the result. Invoke it as $amicited-watermarks in Codex or /amicited-watermarks in Claude Code.

Watermarking and AI visibility

This repository owns AmICited's open-source watermarking layer: preparing and auditing text before publication. The AmICited platform owns the measurement loop after publication—tracking brand mentions, citations, competitor visibility, and source performance across leading AI search engines. Together they support a practical workflow: publish cleaner, reviewed content, then measure whether AI systems discover and cite it.

CLI

Read text from standard input with -, or pass a UTF-8 file path:

printf 'hello\u200bworld\n' | amicited watermark inspect -
amicited watermark verify article.txt
amicited watermark rewrite article.txt
amicited watermark rewrite article.txt --normalization nfc
amicited watermark rewrite article.txt --map-confusables --strip-semantic-format
amicited watermark capabilities

The semantic execution provider is selected with --provider. api is the default and uses a LangChain-supported model:

export OPENAI_API_KEY="..."
amicited watermark rewrite article.txt --provider api --model openai:gpt-5-mini

export ANTHROPIC_API_KEY="..."
amicited watermark rewrite article.txt --provider api --model anthropic:claude-haiku-4-5

export GOOGLE_API_KEY="..."
amicited watermark rewrite article.txt --provider api --model google_genai:gemini-2.5-flash

Alternatively, use an existing authenticated Codex or Claude Code CLI session. The CLI model is optional; omit it to use that tool's configured default:

amicited watermark rewrite article.txt --provider codex
amicited watermark rewrite article.txt --provider codex --model MODEL

amicited watermark rewrite article.txt --provider claude
amicited watermark rewrite article.txt --provider claude --model sonnet

Semantic rewriting is paragraph-oriented. AmICited preserves blank-line separators, splits an overlong paragraph at a preferred sentence boundary or a hard word limit, rewrites the resulting passages concurrently, and reassembles them in source order. Defaults are 180 words per request and four concurrent requests. DIPPER-style lexical and order diversity targets are included in every provider prompt and can be configured explicitly:

amicited watermark rewrite article.txt --provider codex \
  --max-chunk-words 180 --max-concurrency 4 \
  --lexical-diversity 60 --order-diversity 40

These diversity values guide a general model through instructions; they are not the learned control tokens of the original DIPPER model. --temperature sets API-provider sampling only. Codex and Claude use their selected CLI model's generation settings.

Codex and Claude run non-interactively and receive each bounded protected passage directly through standard input. Codex uses an ephemeral read-only sandbox and a temporary final-response file. Claude uses safe mode, no session persistence, and no built-in tools. No original source path is exposed, and the source remains unchanged.

Codex and Claude activity is streamed live to standard error by default, while the final AmICited JSON report is written to standard output. This keeps report redirection machine-readable without hiding provider progress:

amicited watermark rewrite article.txt --provider codex > report.json
# Writes article_dewatermarked.txt and records it in report.json.

Transformation reports exclude source text, transformed text, diff bodies, and finding context by default. The report contains "content_included": false, checksums, output paths, statuses, and diagnostics without copying a private article into JSON. Use --include-content only when a trusted consumer needs the text embedded in the report:

amicited watermark rewrite article.txt --include-content > report-with-text.json

Provider output can include the submitted prompt. Use --no-stream when the terminal output may be recorded or when silent automation is required.

File transformations write a sibling output automatically and preserve the source: article.md becomes article_dewatermarked.md, and article.txt becomes article_dewatermarked.txt. Choose another destination with -o or --output. Existing output files are refused unless --overwrite is explicit:

amicited watermark rewrite article.md --provider codex -o revised.md
amicited watermark rewrite article.md --provider codex --overwrite

Standard-input transformations remain in memory unless --output is supplied, because stdin has no source filename from which to derive a sibling path.

Use --cli-timeout SECONDS to bound either CLI. A missing executable fails before the source file is read. Authentication failures, exhausted usage, timeouts, invalid responses, empty responses, and general CLI failures use stable error categories without copying provider output into the report.

An unqualified model can be paired with --model-provider. A custom OpenAI-compatible endpoint can be selected with --base-url. The CLI validates known provider credentials before reading a file or initializing a request and returns exit code 4 for configuration or model-processing failures. API keys are read from the provider's environment variable and never included in reports.

Implemented commands emit versioned JSON. rewrite and remove transform an in-memory copy, record every deterministic change, run verification before and after the transformation, and never overwrite the source file. Written-output metadata includes the destination, UTF-8 byte and character counts, and SHA-256 checksum.

Python SDK

The SDK never guesses whether a string is text or a path:

from amicited import watermark

report = watermark.rewrite(watermark.WatermarkInput.text("hello\u200bworld"))
print(report.transformed_text)
print(report.to_json())
# Explicit opt-in when serialized JSON must contain text and diff bodies:
print(report.to_json(include_content=True))

semantic_report = watermark.rewrite(
    watermark.WatermarkInput.text("Text to rewrite."),
    model="openai:gpt-5-mini",
    max_chunk_words=180,
    max_concurrency=4,
    lexical_diversity=60,
    order_diversity=40,
)

codex_report = watermark.rewrite(
    watermark.WatermarkInput.text("Text to rewrite."),
    provider=watermark.SemanticProvider.CODEX,
    progress_callback=lambda text: print(text, end=""),
)

Python report objects retain transformed_text in memory. Their to_dict() and to_json() methods redact content by default. Protected-span failures expose content-free diagnostics such as expected/found counts, missing, duplicate or unexpected placeholder IDs, the first mismatch, reordering, and malformed-token counts. A failed rewrite never creates the adjacent output candidate.

The SDK does not stream unless progress_callback is supplied.

watermark.Watermark accepts an ordered sequence of TextWatermarkLayer subclasses. Every layer implements inspect, verify, rewrite, and capability. Inspection and verification run all layers in order. Rewrite and remove feed each layer's output into the next layer and return the individual layer results as well as the aggregate report.

The default order is:

1. HiddenUnicodeLayer
2. BidiControlLayer
3. UnicodeTagLayer
4. ExoticSpaceLayer
5. ConfusableLayer
6. UnicodeNormalizationLayer
7. WhitespacePatternLayer

When an API model is explicitly supplied or the codex/claude provider is selected, an eighth SemanticRewriteLayer runs after the deterministic layers. Its three backends implement the same execution interface; the API backend uses LangChain's provider-neutral init_chat_model. The layer protects the complete document, partitions it into paragraph-level passages bounded by the word limit, and invokes up to max_concurrency provider requests in parallel. Citations, URLs, quotations, numbers, frontmatter, and code are immutable placeholders. Protected-only passages are not transmitted. Restoration succeeds only if every placeholder is returned exactly once and in order. A provider error or protected-span violation preserves the current text and produces a failed transformation.

Context-sensitive joiners, variation selectors, valid emoji tag sequences, and balanced bidi controls are reported but preserved. Exotic spaces are mapped one-for-one; whitespace is never globally collapsed. Confusable mapping, semantic-format stripping, and NFC/NFKC normalization require explicit options. Potentially lossy changes remain visible in the structured change list.

Semantic rewriting is non-deterministic and potentially lossy. Its verification result is always unverifiable: paraphrasing is not a statistical-watermark detector, does not prove removal, and does not establish human authorship. The selected model and provider, external-processing flag, protected-span status, meaning risk, chunk count, word/concurrency limits, diversity targets, changes, and limitations are present in the structured report.

Contributors

Yasha Boroumand Creator and maintainer

Community contributions are welcome. See everyone who has helped on the GitHub contributors page.

Development

uv sync
uv run pytest
uv run amicited --help
uv build

Release

Releases are published to PyPI through GitHub Actions and PyPI Trusted Publishing. Set the version in pyproject.toml, commit it, then push a matching tag such as v0.1.0. The release workflow rejects tags that do not exactly match the package version, runs the full quality suite, builds from the source distribution, and publishes only after those checks pass.