Pangram/cliunofficial

AI detection,
without leaving
your terminal.

An unofficial CLI, interactive TUI, and typed stdio MCP server for Pangram. Built for shell pipelines, human review, and agent workflows.

Install on macOS or Linux
$ curl -fsSL https://pangram.micr.dev/install | sh

Signed installers are available for macOS, Linux, and Windows.

pangram - local sessionready
Pangram> [Analyze]  Active  History  Settings
AnalyzeChecks> [x] AI detection - available  [ ] Plagiarism - available  [ ] Both - availableInput> [x] Text - available
Text composer

Paste or type text here

Inspector[ ] Public link: off[ ] Manual save: offWords: 32Billable estimate: 1 unit[Enter] Submit
[Tab] Next  [Shift+Tab] Previous  [?] Help  [Enter] Quit
01Rust 1.88+single native runtime
02JSON-firstcomposable by default
03CLI · TUI · MCPone shared core
04MIT licensedsource available
Illustrated greenhouse workspace with a person at a computer
$ pangramstdout result.jsonPANGRAM / CLI / 001

One terminal client. Three ways to work.

Run detection from a shell, open the interactive terminal interface, or expose the same analysis layer to agents through a typed stdio MCP server.

pangram detectliteral text and stdin
pangramall-TTY launch opens the TUI
pangram mcptyped agent tools over stdio
Read the quickstart →

A command surface that
stays composable.

Structured output, explicit inputs, and predictable separation between results and diagnostics. AI detection, text plagiarism, and combined text analysis are compiled today.

01pangram detect

Analyze text directly.

Pass literal text, -, or piped stdin. JSON is canonical outside the TUI.

02pangram bulk

Handle asynchronous work.

Submit, inspect status, wait, and fetch results without hand-rolling polling logic.

03pangram history

Keep local records, optionally.

History is off by default. Enable it for local search, export, deletion, and reruns.

04pangram mcp

Give agents typed tools.

Use a stdio MCP server with explicit safeguards for billing, files, history, and mutations.

Built for shell pipelines,
people, and agents.

The interface changes. The analysis behavior does not. Each adapter sits on the same Rust module for HTTP, polling, normalization, retries, and task state.

{ }

JSON is canonical

Machine-readable output is the default for noninteractive use.

01
Aa

Readable when needed

Switch to pretty, Markdown, or TOON projections.

02
↳

Clean stream boundaries

Primary results go to stdout; diagnostics stay on stderr.

03
!

Structured failures

Errors carry codes, categories, retryability, and recovery actions.

04
DB

Opt-in local history

Store records locally only after explicit consent.

05
API

Documented REST only

No browser sessions, private dashboard routes, or scraping.

06
TTY

Responsive TUI

Analyze, Active, History, and Settings in one focused interface.

07
MCP

Agent-native contracts

Typed tools and version-matched embedded guidance.

08
Illustrated modern railway station at dusk
quickstart
$ pangram auth$ pangram detect 'Text to analyze'$ printf 'Text' | pangram

From API key
to result in a few commands.

Run guided setup with pangram auth. PANGRAM_API_KEY can override the local credential for ephemeral environments.

One core. Three entry points.

Use the surface that fits the moment without inventing separate behavior for each environment.

Illustrated modern library interior
01

Shell

Scriptable by default.

Pipe text in, receive structured output, and compose it with standard Unix tools.

printf 'Text' | pangram
Illustrated greenhouse computer workspace
02

TUI

Interactive when you are.

Launch with no arguments in an all-TTY session to open Analyze, Active, History, and Settings.

pangram
Illustrated computer lab in a wooded landscape
03

MCP

Typed tools for agents.

Run a stdio server with explicit boundaries for file access, billing, history, and mutations.

pangram mcp

Commands available
right now.

Detection, plagiarism, combined analysis, bulk workflows, local history, the TUI, and MCP share one release.

01pangram auth

Configure and inspect API-key authentication.

02pangram config

List, get, set, and locate non-secret configuration.

03pangram doctor

Run local, non-billable diagnostics.

04pangram detect

Run Pangram 4 AI text detection.

05pangram bulk

Submit, inspect, wait, and fetch async results.

06pangram task

Inspect or wait for one Pangram text task.

07pangram history

Search, export, delete, clear, and rerun local records.

08pangram mcp

Run or install the typed stdio MCP server.

09pangram agent

Print the embedded agent usage guide.

10pangram skills

List and load version-matched embedded skills.

11pangram plagiarism

Check text for plagiarism.

12pangram analyze

Run detection and plagiarism together.

13pangram completions

Generate shell completion scripts.

14pangram update

Report signed-update availability.

Release channels
native installersLinux, macOS, and Windows
package managersnpm, Homebrew, and Scoop

One analysis core.
No duplicate behavior.

The Rust analysis module owns Pangram HTTP behavior, polling, normalization, retries, and task state. CLI, TUI, and MCP are adapters over it.

UpstreamPangram RESTAPI-key authenticated
Shared coreAnalysis moduleHTTP · polling · normalization · retries · task state
01CLIJSON / pretty
02TUIinteractive review
03MCPtyped tools
same contractsseparate adapters

History stays off
until you turn it on.

When enabled, local history stores canonical inputs, results, task identity, and searchable text in SQLite until you delete them. Public Pangram dashboard links are also disabled by default and requested per submission.

Read the privacy notes →
History disabled by defaultPublic links disabled by defaultDestructive actions separately gated
Illustrated university archive with rows of catalog drawers
history.enabledfalseexplicit consent required

Questions,
answered.

The important boundaries, without the marketing fog.

Project docs

No. It is an independent, unofficial client and is not affiliated with, endorsed by, or connected to Pangram Labs, Inc.

MIT LICENSED

Bring Pangram into the terminal.

Install the CLI, add your Pangram API key, and choose the shell, TUI, or MCP workflow that fits the job.

Illustrated computer lab glowing in a wooded landscape
$ pangram