Introducing Codex
An interface to the complete reference corpus for every program, utility, and function that powers your infrastructure. Surface authoritative documentation instantly. Structured. Indexed. Canonical.
Backed by mandō
We exist in an age where critical documentation is scattered across wikis, forums, and outdated blog posts. The canonical reference — precise, structured, authoritative — has been displaced by noise. At Codex, we believe that the path to systems mastery does not begin with a search engine. It begins with a system of record.
Codex provides an interface to the complete reference manuals for every command, system call, library function, device driver, file format, and kernel routine in your stack. Not summaries. Not opinions. The definitive text. We have built a hierarchical document architecture spanning nine distinct sections of knowledge — a taxonomy refined by decades of systems thought, now delivered at the speed your teams require.
This is not another documentation platform. This is the primary source.
Codex organises the world's systems knowledge into nine canonical sections. Each section is a curated domain of expertise, indexed, cross-referenced, and governed by strict editorial standards.
Executable programs and shell commands. The operational surface area of every system. Every utility your team invokes daily, documented to the flag level.
Functions provided by the kernel itself. The raw interface between userland and kernel space. Syscall signatures, return values, error codes — all canonical.
Functions within program libraries. The building blocks of systems software. C standard library, POSIX interfaces, and beyond — structured and searchable.
Device nodes and special files. The interface layer between software and hardware. Complete device documentation, usually found in /dev.
File formats and conventions. Configuration schemas, data structures, protocol specifications. Know exactly how every system file is structured.
Games and screensavers. Even recreational computation deserves rigorous documentation. Complete reference for every interactive program.
Macro packages, conventions, and overviews. The connective tissue of systems knowledge — protocols, character sets, filesystem hierarchies.
System administration commands. Root-level operations, daemon management, and infrastructure maintenance — the tools of the operator class.
Kernel internals and routines. The deepest layer of systems documentation. For those who operate at the boundary of hardware and software.
Every feature maps to a fundamental operation in knowledge retrieval. No abstraction layer. No compromise.
codex ls — Retrieve the canonical reference for any command, function, or utility by
name. One query, one authoritative result. The default path to knowledge.
codex 3 printf — When a name exists across multiple domains, scope your query to a
specific section. Disambiguate between the command and the library call with precision.
codex -k "pattern" — Search across all short descriptions and page names using
keyword matching. Surface every relevant document in the corpus with a single query.
codex -f command — Retrieve the short-form description for any page in the archive.
Rapid identification without full page retrieval. Optimised for reconnaissance workflows.
codex -a intro — Display all matching pages across every section, in succession.
When one result is insufficient, surface the complete set. Leave nothing undiscovered.
codex -w command — Reveal the precise filesystem location of any source document.
Map your knowledge infrastructure. Understand where truth resides.
codex -K "term" — Brute-force search across the complete text of every page in the
archive. When keyword search isn't enough, go deeper. Search the source itself.
codex --regex "pat.*" — Elevate your queries with regular expression and wildcard
pattern matching. Surface documents that simple string matching would miss entirely.
From query to canonical knowledge in four stages.
Specify the subject of inquiry — a command, function, or topic — optionally scoped to a section. The query interface accepts natural identifiers, section-qualified names, and pattern expressions.
Codex traverses the configured MANPATH hierarchy, resolving the query against installed documentation sources. Multiple paths are evaluated in priority order, respecting locale and section precedence.
The source document is processed through the formatting pipeline — preprocessors, nroff/groff rendering, and encoding normalisation. The raw reference material is transformed into a readable, structured presentation.
The formatted output is delivered through a configurable pager with prompt customisation, search capabilities, and navigation controls. Consume knowledge at your own pace, with full control over the reading experience.
Experience the Codex retrieval engine. Enter a query below and observe the structured results from our indexed corpus.
Enter a query to search the Codex archive.
A live demonstration of the Codex document viewer. Select a page from the corpus and observe the formatted reference output, rendered with full structural fidelity.
NAME
ls - list directory contents
SYNOPSIS
ls [OPTION]... [FILE]...
DESCRIPTION
List information about the FILEs (the current
directory by default). Sort entries
alphabetically if none of -cftuvSUX nor
--sort is specified.
Mandatory arguments to long options are
mandatory for short options too.
-a, --all
do not ignore entries starting with .
-l use a long listing format
-h, --human-readable
with -l, print sizes like 1K 234M 2G etc.
-R, --recursive
list subdirectories recursively
-t sort by time, newest first
Observe Codex in its native environment.
Codex provides granular control over every aspect of the reading experience. Configure pager behaviour, locale settings, and search path resolution to match your organisational requirements.
codex -P less
Choose the output renderer. Ship with less, cat, or any custom pager. Control how knowledge is consumed at the team level.
codex -L de_DE
Deliver documentation in the reader's native language. Multi-locale support across your entire knowledge corpus.
MANPATH=/custom/docs
Define custom search hierarchies. Overlay proprietary documentation alongside system references. Build a unified knowledge graph.
codex -S 3:2:1
Prioritise which knowledge domains are searched first. Developers get library calls; operators get admin commands. Role-based retrieval.
"We replaced our entire internal documentation wiki with Codex. The section-scoped lookup alone saved our SRE team hundreds of hours of context switching. This is first-principles knowledge retrieval."
Dr. Elaine Marchetti VP of Platform Engineering, NovaSystems
"The full-text corpus search is a genuine paradigm shift. Before Codex, finding where TCP_NODELAY is documented meant crawling forums. Now it's one query. Category-defining product."
Ravi Chandrasekaran Principal Systems Architect, Meridian Computing
"I shipped Codex to 400 engineers on day one. The MANPATH configuration meant we could layer our internal docs alongside the canonical references. The compound value is extraordinary."
Katrina Ohlsson CTO, Substrate Labs
"As a solo developer, Codex's apropos search is my north star. I don't need to know the exact command — I describe what I want, and the knowledge surfaces. This is how documentation should work."
Marcus Yeung Independent Software Consultant
Knowledge retrieval, priced for every scale of operation.
Online documentation is scattered, inconsistent, and frequently out of date. Codex provides a single, structured interface to the canonical reference corpus, indexed by section, searchable by keyword, regex, or full-text, and delivered through a configurable pager optimised for deep reading. This is the difference between browsing a library and owning one.
Absolutely. Using MANPATH configuration, Department and Institute tier customers can overlay internal documentation alongside the canonical corpus. Your team accesses proprietary and public references through a single interface, with no context switching.
Codex follows a configurable section precedence order by default (1, n, l, 8, 3, 0, 2, 5, 4, 9, 6, 7). You can override this globally or per-query using section-scoped lookup. The -a flag retrieves all matching pages across every section, ensuring comprehensive coverage.
Yes. Codex operates entirely against locally installed reference material. No network connectivity is required for retrieval operations. Institute tier includes on-premise deployment with a dedicated knowledge architect for configuration.
Apropos (-k) searches only page names and short descriptions — ideal for known-item retrieval. Full-text corpus search (-K) examines the complete source of every page in the archive. It's a brute-force operation designed for when you know a term appears somewhere in the documentation but don't know which page contains it.