Context continuity across AI agents & sessions. Team knowledge collaboration. Industrial-grade AI delivery.
AI coding agents are powerful, but they forget everything between sessions, lose all context when you switch agents, and have no way to share knowledge across a team. Every conversation starts from zero.
WITHOUT fcontext
Session 1 (Copilot): "Build the auth module"
Session 2 (Claude): "What auth module? I don't see any context."
Session 3 (Cursor): "Can you explain the project first?"
Teammate's Agent: "I have no idea what your team has decided."
Result: Repeated explanations. Inconsistent output. Wasted tokens.
WITH fcontext
Session 1 (Copilot): "Build the auth module" -> saves context
Session 2 (Claude): Reads _README.md + _topics/ -> picks up work
Session 3 (Cursor): Knows the full project, requirements, history
Teammate's Agent: Imports experience pack -> instant onboarding
Result: Continuous context. Consistent quality. Industrial output.
AI delivers results, but you deliver process and experience.
Your expertise — how you approach problems, what patterns you've learned, what pitfalls to avoid — is lost every time a session ends. fcontext captures and persists that experience, so your AI gets smarter with every interaction.
- Cross-session memory — Topics and conclusions survive between conversations
- Cross-agent portability — Switch from Copilot to Claude to Cursor without losing context
- Experience visualization — Your accumulated knowledge becomes a structured, reusable asset
No single agent has all the context to do the job. Real work is distributed.
In production environments, context is fragmented: requirements live in documents, domain knowledge lives in people's heads, architecture decisions live in past conversations. fcontext solves distributed context.
- Team knowledge sync — Export/import experience packs so every team member's agent shares the same domain understanding
- Requirements traceability — Track stories, tasks, bugs with full evolution history — from document to delivery
- Compliance-ready — All data stored locally in
.fcontext/, no cloud dependency, fully offline capable - Industrial-grade delivery — Structured context + requirements tracking + document indexing = consistent, auditable AI output
| Capability | Description | Data Location |
|---|---|---|
| Multi-Agent Support | Works with all mainstream AI coding agents | Agent-native config files |
| Document Indexing | PDF, DOCX, XLSX, PPTX, Keynote, EPUB to Markdown | .fcontext/_cache/ |
| Dynamic Context Building | AI accumulates knowledge topics across sessions | .fcontext/_topics/ |
| Experience Packs | Import/export domain knowledge across projects and teams | .fcontext/_experiences/ |
| Requirements Management | Stories, tasks, bugs with evolution tracking | .fcontext/_requirements/ |
| Workspace Map | Auto-generated project structure overview | .fcontext/_workspace.map |
| Living Project Summary | AI-maintained _README.md, first thing every session reads |
.fcontext/_README.md |
| Offline & Secure | All data local. No cloud. No API keys. No telemetry. | .fcontext/ |
| Agent | Command | Config Format |
|---|---|---|
| GitHub Copilot | fcontext enable copilot |
.github/instructions/*.instructions.md |
| Claude Code | fcontext enable claude |
.claude/rules/*.md |
| Cursor | fcontext enable cursor |
.cursor/rules/*.md |
| Trae | fcontext enable trae |
.trae/rules/*.md |
| OpenCode | fcontext enable opencode |
Uses Claude format |
| OpenClaw | fcontext enable openclaw |
skills/ only |
- Python 3.9+
- pip
pip install fcontextfcontext --version
# fcontext 1.0.0# 1. Initialize in any project
cd your-project
fcontext init
# 2. Activate your AI agent
fcontext enable copilot # or: claude, cursor, trae, opencode, openclaw
# 3. Index your documents
fcontext index docs/
# 4. Check status
fcontext statusThat's it. Your AI agent now reads project context automatically on every session.
Problem: You had a deep debugging session yesterday. Today, a new session knows nothing.
# fcontext automatically persists session knowledge to _topics/
# Next session reads _topics/ and _README.md first
# To see what was saved:
fcontext topic list
fcontext topic show debugging-auth-flowThe new session starts with full context of yesterday's findings.
Problem: You used Cursor for frontend work but need Claude for backend refactoring. Claude has no idea what Cursor did.
# Enable both agents — they share the same .fcontext/ data
fcontext enable cursor
fcontext enable claude
# Both agents read the same _README.md, _topics/, _requirements/
# Context is agent-agnosticProblem: A new developer joins. Their AI has zero project knowledge.
# Team lead exports accumulated knowledge
fcontext export team-knowledge.zip
# New member imports it
fcontext experience import team-knowledge.zip
# Their AI instantly knows: architecture, domain concepts, conventions, pitfalls
fcontext experience listProblem: Product specs are in PDF/DOCX. AI cannot read them.
# Convert to Markdown so any agent can read them
fcontext index specs/product-requirements.pdf
fcontext index contracts/
# AI now references the content directly from _cache/
fcontext statusProblem: Requirements are scattered across documents, Slack, and meetings. AI builds the wrong thing.
# Structure requirements in fcontext
fcontext req add "User authentication via OAuth" -t story
fcontext req add "Support Google and GitHub providers" -t task --parent STORY-001
fcontext req set TASK-001 status in-progress
# AI reads _requirements/ and builds against tracked specs
fcontext req board # Kanban view
fcontext req tree # Hierarchy viewProblem: You've built deep domain knowledge in Project A. Project B needs the same expertise.
# In Project A: export to a git repo
fcontext export git@github.com:team/domain-knowledge.git
# In Project B: import as experience pack
fcontext experience import git@github.com:team/domain-knowledge.git
# Keep it updated
fcontext experience update| Command | Description |
|---|---|
fcontext init |
Initialize .fcontext/ in workspace |
fcontext enable <agent> |
Activate an AI agent |
fcontext enable list |
Show all supported agents and status |
fcontext status |
Show index statistics |
fcontext clean |
Clear cached files |
fcontext reset |
Delete all .fcontext/ data |
| Command | Description |
|---|---|
fcontext index |
Scan and convert all files in workspace |
fcontext index <file> |
Convert a specific file |
fcontext index <dir> |
Convert all files in a directory |
fcontext index -f |
Force re-convert even if up-to-date |
| Command | Description |
|---|---|
fcontext req add "title" -t TYPE |
Add item (roadmap/epic/story/task/bug) |
fcontext req list |
List all items (supports --type, --status filters) |
fcontext req tree |
Hierarchy view |
fcontext req board |
Kanban board by status |
fcontext req show ID |
Item details + changelog |
fcontext req set ID field value |
Update a field |
fcontext req link ID TYPE TARGET |
Link items (supersedes/evolves/relates/blocks) |
fcontext req trace ID |
Follow evolution chain |
fcontext req comment ID "msg" |
Add a comment |
| Command | Description |
|---|---|
fcontext topic list |
List accumulated knowledge topics |
fcontext topic show <name> |
Show topic content |
fcontext topic clean |
Remove empty topic files |
| Command | Description |
|---|---|
fcontext experience list |
Show imported packs |
fcontext experience import <source> |
Import from zip, git URL, or download URL |
fcontext experience remove <name> |
Remove a pack |
fcontext experience update [name] |
Update from original source |
fcontext export <output> |
Export knowledge to zip or git remote |
your-project/
.fcontext/ # All context data (git-tracked)
_README.md # AI-maintained project summary
_workspace.map # Auto-generated structure
_index.json # File index registry
_cache/ # Converted documents (Markdown)
_topics/ # Session knowledge & conclusions
_requirements/ # Stories, tasks, bugs
items.csv # Structured data
_backlog.md # Auto-generated summary
docs/ # Per-item details
_experiences/ # Imported domain knowledge (read-only)
<pack-name>/
_README.md
_cache/
_topics/
.github/instructions/ # Copilot (auto-generated by fcontext enable)
.claude/rules/ # Claude (auto-generated)
.cursor/rules/ # Cursor (auto-generated)
.trae/rules/ # Trae (auto-generated)
Each AI agent gets instructions in its native format. The instructions teach the agent to:
- Read
.fcontext/_README.mdfirst to understand the project - Check
_cache/before trying to read binary files - Use
fcontext reqcommands for requirements (never parse CSV manually) - Save important conclusions to
_topics/for future sessions - Read
_experiences/for imported domain knowledge
We welcome contributions! Here's how to get started:
# 1. Fork and clone the repository
git clone https://github.com/<your-fork>/fcontext.git
cd fcontext
# 2. Install in development mode
pip install -e ".[test]"
# 3. Run the test suite
pytest tests/
# 213 tests should pass
# 4. Understand the structure
# fcontext/
# cli.py # CLI entry point & argument parsing
# init.py # Workspace initialization & agent configs
# indexer.py # File scanning, conversion, text copy
# experience.py # Experience pack import/export/update
# requirements.py # Requirements CRUD & reporting
# topics.py # Topic management
# workspace_map.py # Project structure generation
# 5. Make your changes, add tests, verify
pytest tests/ --tb=short
# 6. Create a PR (never push directly to main)
git checkout -b feat/your-feature
git commit -m "feat: description"
git push -u origin feat/your-feature- Every new feature needs tests
- Keep
from __future__ import annotationsin all modules (Python 3.9 compat) - CLI commands should give clear error messages when
.fcontext/is not initialized - Experience packs under
_experiences/are always read-only
This project is licensed under the Apache License 2.0.