AI Context Builder 100% local

AI Context Builder for Codex & Claude

Moving from Codex to Claude? Give your next agent your current project: “Get familiar with this code before we continue.”

100% local · No signup · Free: one pack per day, 25 files, 20,000 estimated tokens · 10 MB per file

AI Context Builder for Local Files
Token budget and saved profiles
Give your next agent the project files it needs to get familiar with your code.
Source files
Documents, source code, configs, or a project folder
No files
0Files
0Context tokens
ReadyTarget fit
0Secret warnings
No context No files
context.md
Editable LLM-ready Markdown handoff

0
local conversions counted worldwide Files stay local during preparation. Usage events contain no file names, contents or task text.

AI context builder

Help Claude get familiar with your current project.

Combine documents, source code, configuration files, or a local project folder in one structured handoff for ChatGPT, Claude, Codex, Gemini, or another LLM. MarkDone preserves relative folder paths, keeps source boundaries visible, estimates tokens, and detects common secrets without uploading your files.

Agent-to-agent handoffsMove from Codex to Claude with your current project files. Give the new agent a clear starting point with relative paths and visible source boundaries.
Secret-awareDetect and optionally redact common keys, tokens, private keys, and password-like values before pasting.
Private by designPDF extraction, DOCX parsing, scanning, and token counting run in the browser.

Give your next coding agent the context it needs

  1. Drop supported files or choose Add folder and review its contents.
  2. Choose Switch agent to ask the new agent to get familiar with the project. Add your current goal or saved project notes if useful.
  3. Review the selected sources, token estimate and secret redaction.
  4. Copy the Markdown into the new agent or download context.md and ask it to read the handoff before continuing.

Select and review the sources you include

Include or exclude each file before sharing. The generated Markdown keeps file names, metadata, token estimates, and clear begin/end markers so you can review exactly what an LLM will receive.

Keep secrets out of AI prompts

MarkDone checks common API keys, bearer tokens, private keys, JWTs, cloud keys, and password-like assignments. Optional redaction replaces detected values locally before they enter context.md.

Check the token budget before sharing

Choose a planning target and compare it with the estimated context size. If the pack is too large, exclude low-value sources or split the work into smaller, task-specific packs.

Start with your project, then continue with a new agent

Moving from Codex to Claude starts with a simple introduction: “This is my current project. Get familiar with it before we continue.” Bundle the relevant code and project notes into portable Markdown, then ask the new agent to explain what it understands and what context is missing. For the complete workflow, read how to build an LLM context pack from local files.

Context pack vs CLAUDE.md, AGENTS.md and project memory

MarkDone creates a portable file-based handoff for the task in front of you. It does not generate persistent tool-specific instruction files or replace the native project memory of a coding agent. See the Claude and Codex project-context guide for the practical differences.

Can I switch from Codex to Claude with this tool?

Yes. Add your current project folder, select the relevant files, and choose Switch agent. Copy or download context.md and give it to Claude with a request to get familiar with the architecture, key files and current state before continuing. You can include saved project notes too. MarkDone packages the files and notes you provide; it does not automatically retrieve chat history or the previous agent's memory.

What is an AI context builder?

An AI context builder prepares the source material an LLM needs. MarkDone extracts readable text, preserves file names and source boundaries, checks the token budget, warns about common secrets, and creates one portable context.md file.

What should a context.md file contain?

A useful context.md contains a clear task, a source summary, file metadata, warnings, and one labeled section per source so an LLM can distinguish and compare the evidence.

Are my files uploaded?

No. MarkDone extracts text, counts tokens, scans for secrets, and builds context.md inside your browser. Your source files are not uploaded to MarkDone or to a model provider.

Can I combine PDF, DOCX, Markdown, JSON, TOON, YAML and CSV?

Yes. You can combine readable PDF and DOCX files with Markdown, plain text, JSON, TOON, YAML, YML, and CSV sources in one context.md. TOON is preserved as readable structured text. Scanned PDFs require OCR and are not supported yet.

Can I add a whole folder?

Yes. Click the main drop area, choose Add folder, or drag and drop a folder to review supported files before import. MarkDone preserves relative paths, leaves files from common generated folders such as node_modules and dist unselected by default, and adds an included file tree to context.md. Free includes one context pack with up to 25 files per day.

Can I include Swift and other source code?

Yes. MarkDone supports common Apple, web, backend, systems, shell, and project file formats, including Swift, JavaScript, TypeScript, Python, Go, Rust, Java, Kotlin, C, C++, shell scripts, Dockerfile, Makefile, and common configuration files. Sensitive configs, lockfiles, generated folders, and minified output remain visible but start unchecked. Binary files are not read as source text.

Does the project importer respect .gitignore?

Yes. MarkDone reads root and nested .gitignore files locally, applies common Git ignore patterns including wildcards, directory rules, anchored paths, and negation rules, and leaves matched files visible but unchecked in the review. You can still include an ignored file manually.

How do I use context.md with Claude or Codex?

Enter your task or choose a starter task, review the generated Markdown, then copy it into a chat or download context.md for a project handoff. The plain Markdown format also works with ChatGPT, Gemini, coding agents, and local models.

What happens when the context pack is too large?

MarkDone compares the estimated token count with the selected planning budget. If the pack is too large, exclude low-value files or split the task into smaller context packs before sharing it.

Does MarkDone generate CLAUDE.md or AGENTS.md?

No. MarkDone generates a portable context.md handoff from the files you select. It does not create tool-specific project instruction files such as CLAUDE.md, AGENTS.md, or GEMINI.md.