Bora

Bora — Brazilian Portuguese slang for let's go

Bora is a CLI that keeps your AI collaborator oriented across sessions, models, and projects.

Whether you're a developer building software with an AI coding agent, or a writer using AI to research and plan a manuscript, bora solves the same two problems:

1. Context decay. Every new AI chat starts from zero. You re-explain the project, often inconsistently, and details get lost.
2. Model drift. Switching between Claude, GPT, Gemini, or a local model means starting the briefing over.

Bora fixes this by maintaining a small, structured set of Markdown files inside your project. Any model — any tool — can read them to get oriented in seconds. The files travel with your project, stay in version control, and are always the authoritative source of what's happening and why.

Bora 0.4.5 ships two isolated profiles:

• dev — for software projects: hierarchical tickets, a dated Requirements spec, per-project status dashboards, and AI tool skill integration.
• write — for writing projects: chapter scaffolding, research interaction logs, story context, and summary generation.

Contents

• Installation
• Profiles
• For developers
  • How it works
  • Quick start
  • Commands
  • AI tool skills
  • Conventions
• For writers
  • How it works
  • Quick start
  • Commands
  • Obsidian integration
  • Conventions
• Working across models
• Upgrading
• Contributing

Installation

The recommended way to install Python CLIs globally is pipx, which gives bora its own isolated environment and puts it on your PATH without touching your system Python.

Install pipx (if you don't have it)

# macOS
brew install pipx
pipx ensurepath

# Linux / WSL
python3 -m pip install --user pipx
python3 -m pipx ensurepath

# Windows (PowerShell)
python -m pip install --user pipx
python -m pipx ensurepath

After running pipx ensurepath, open a new terminal so the updated PATH takes effect.

Install bora

# From PyPI (once published):
pipx install bora

# From GitHub:
pipx install git+https://github.com/tonyknight/Bora.git

Verify the install:

bora --version

Upgrade

pipx upgrade bora

Uninstall

pipx uninstall bora

Alternative: pip --user

If you prefer not to use pipx:

pip install --user bora

If bora isn't found after install, your user binary directory (~/.local/bin on Linux/macOS) may not be on your PATH. Add it:

# ~/.bashrc or ~/.zshrc
export PATH="$HOME/.local/bin:$PATH"

Development install (from source)

git clone https://github.com/tonyknight/Bora.git
cd Bora
pipx install -e .

Changes to the source take effect immediately — no reinstall needed.

Profiles

Every bora project has a profile stored in .bora/profile.json. The profile determines which commands are available and which appear in --help.

Bora enforces profile isolation: running a dev command in a write project (or vice versa) exits with a clear error. This keeps the two workflows from interfering with each other.

If you run a command in a project that has no profile yet (for example, a project created before 0.3.0), bora prompts you to choose one and writes the file before continuing.

For developers

How it works (dev)

bora dev init <project_path> scaffolds one project under a hierarchical path. Multiple projects may coexist under docs/ai/. There is no active-project pointer — every command takes an explicit <project_path>.

Example: bora dev init "QromaCore/Hamburg/Gallery Refactor" on 2026-08-14 yields:

AGENTS.md              ← AI agent instructions (repo root only)
docs/
  ai/
    QromaCore/
      Hamburg/
        Gallery Refactor/
          (2026-08-14) Gallery Refactor.md              ← project briefing
          (2026-08-14) Gallery Refactor Requirements.md ← architecture + spec
          Status.md                                     ← auto-generated dashboard
          tickets/
            .gitkeep
.bora/
  profile.json         ← profile lock (dev)

Deeper paths are valid too: bora dev init Acme/Platform/Auth/OAuth Refresh (depth 4). {ProjectName} is the last segment.

The dated briefing describes what is being built and why. After the human and agent discuss architecture, they fill the sibling Requirements file — that document is the per-project spec and the source of tickets (from its Tasks Breakdown). Status.md is regenerated from that project's tickets any time you run bora dev status <project_path>. Never hand-edit it. Before marking a ticket done or committing, follow Commit criteria in the Requirements file (see Dev conventions).

Each ticket is a Markdown file under docs/ai/<path>/tickets/ with YAML frontmatter for machine-readable state (status, priority, depends_on, subtasks) and a free-form body for human-readable context (description, acceptance criteria, notes). Ticket IDs are unique per-project, not repo-global.

An AI agent reads root AGENTS.md first (scope guardrail and Requirements workflow), then the human-referenced project briefing. It can run bora dev ticket set <project_path> …, bora dev ticket note <project_path> …, and bora dev lint <project_path> directly — the whole loop works from inside a model's shell.

Quick start (dev)

# Start in your repo (existing or new)
cd /path/to/my-codebase

# Scaffold a hierarchical project (quoted form; spaces are one segment)
bora dev init "QromaCore/Hamburg/Gallery Refactor" --tags Codebase,"Release Train",Project

# Same path, shell-escaped, without tags (frontmatter gets hierarchy only)
bora dev init QromaCore/Hamburg/Gallery\ Refactor

# Fill in what you're building
$EDITOR "docs/ai/QromaCore/Hamburg/Gallery Refactor/(2026-08-14) Gallery Refactor.md"

# Discuss architecture with your agent, then fill the Requirements file.
# Do not create tickets until the Tasks Breakdown is agreed.

# Create a ticket from the Requirements Tasks Breakdown
bora dev ticket new "QromaCore/Hamburg/Gallery Refactor" "Implement login flow" --priority high

# Mark it in-progress (fuzzy ID match — no need to type the full ID)
bora dev ticket set "QromaCore/Hamburg/Gallery Refactor" 01 status in-progress

# Add a note as work progresses
bora dev ticket note "QromaCore/Hamburg/Gallery Refactor" 01 "Auth service wired up, still need session handling"

# Regenerate the per-project status dashboard
bora dev status "QromaCore/Hamburg/Gallery Refactor"

# Brief a fresh model session (pipe to clipboard)
bora dev context "QromaCore/Hamburg/Gallery Refactor" | pbcopy

Missing <project_path> is an error. There is no active-project fallback.

Dev commands

Project initialisation

Removed in 0.4.5

• bora dev project — versioning and docs/ai/Projects/ archival are gone. Hierarchy and optional tags are set at init: bora dev init <project_path> --tags ….
• bora dev decision — record decisions in the project's Requirements file (typically under Architecture or Open questions).

Tickets

Every ticket command requires <project_path> first. Missing it is an error.

Project state

AI tool skills

AI tool skills

Claude Code, OpenCode, and other agentic tools support skills — directories containing a SKILL.md that the agent loads when its description matches the current task. Bora ships a skill that tells the agent how to use the CLI and maintain the hierarchical docs/ai/ files correctly.

Install the skill after bora dev init <project_path>:

# User-level — available in every project
bora dev skill install claude
bora dev skill install all

# Project-level — ships inside this repo
bora dev skill install claude --project

Show what's installed where:

bora dev skill list

Remove:

bora dev skill uninstall claude          # user-level
bora dev skill uninstall claude --project  # this repo only

The uninstall command only removes a SKILL.md whose frontmatter identifies it as bora's — it won't delete a skill it didn't create. Pass --force to override.

Dev conventions

• Ticket IDs are generated, not chosen. The format is YYYYMMDD-NN-slug. Never rename ticket files — the ID is the source of truth for depends_on and parent references. IDs are unique per-project, not repo-global.
• Status.md is per-project and auto-generated. Never hand-edit it. Update tickets and run bora dev status <project_path>. There is no root docs/ai/Status.md aggregation.
• Commit criteria gate done work and git commits. Before marking a ticket or subtask done, and before any git commit, satisfy the Commit criteria section of that project's Requirements file: the subtask's completion tests pass, the change meets the requirement, and build/tests pass (including platform builds such as macOS/iOS when that is the target). Commit message format: {task name}: {summary of what was done}. Bora does not create git commits; this is agent workflow.
• AGENTS.md is root-only. It contains the scope guardrail and the discuss-architecture → write Requirements → create tickets workflow. Init writes it once; it is not overwritten unless you pass --force.
• Subtasks live in two places by design. Major subtasks appear in frontmatter (subtasks: list) — they're queryable and visible in Status.md. Small subtasks are Markdown checkboxes in the ticket body — they're counted but not individually tracked.
• Decisions live in the Requirements file. After agreeing with the human, edit Architecture or Open questions in the dated Requirements file. There is no decision command.
• No migration for flat layouts. 0.4.5 does not convert docs/ai/Project.md trees. New work uses hierarchical docs/ai/<path>/ projects.

For writers

How it works (write)

bora write init scaffolds a writing project:

AGENTS.md              ← AI agent instructions (role, boundaries, workflow rules)
doc/ai/
  Project.md           ← story overview: premise, plot, characters, worldbuilding
Summary.md             ← latest AI-generated context briefing (ephemeral)
Summary/               ← archive of previous summaries
.bora/
  profile.json         ← profile lock (write)

As you create chapters, bora adds:

Chapters/
  Chapter 001 - The Arrival/
    001 - The Arrival.md       ← manuscript (author-only, agents never write here)
    001 - ChapterProject.md    ← planning: beats, arcs, pacing, agent notes
    001 - Research.md          ← research log: AI interactions, sources, verification

The core workflow is:

1. Write your chapter notes and research questions in ChapterProject.md.
2. Ask your AI to research topics. The agent logs interactions in Research.md.
3. Run bora write status to compile everything into a structured briefing.
4. Paste the briefing into a new chat with your AI. Ask it to generate an updated Summary.md.
5. Save the response as Summary.md. Next time you run bora write status, it's archived and a fresh one is generated.

The AI never touches your manuscript. The creative work stays yours.

Quick start (write)

# Create and enter your project directory
mkdir my-novel && cd my-novel

# Scaffold the write project
bora write init

# Set up your story context
$EDITOR AGENTS.md              # tell your AI its role and boundaries for this story
$EDITOR doc/ai/Project.md      # add your premise, characters, plot structure

# Create your first chapter
bora write chapter "The Arrival"

# Fill in the chapter plan
$EDITOR "Chapters/Chapter 001 - The Arrival/001 - The Arrival ChapterProject.md"

# Add a second chapter
bora write chapter "The Conflict"

# Compile a briefing for your AI model
bora write status

# Paste the output into Claude / ChatGPT / your model of choice
# Ask: "Based on this project context, generate an updated Summary.md"
# Save the response as Summary.md

Write commands

Project initialisation

Chapters

Project status

Skills

Obsidian integration

If you use Obsidian as your writing environment, bora can install a vault-aware prompt that orients your AI to the project's structure:

bora write skill install obsidian

This creates .obsidian/plugins/bora-writer/ with:

• SKILL.md — a prompt template describing the vault structure, the chapter layout, the Research.md format, and the rule that agents never write to manuscript files. Paste its contents into your AI model's system prompt or context window.
• manifest.json — Obsidian community plugin metadata.
• README.md — setup instructions.

To remove it:

bora write skill uninstall obsidian

Write conventions

• Agents never write to manuscript files. The 001 - Chapter Name.md files are author-only. If an agent modifies one, treat that as a mistake — the AGENTS.md template instructs it not to.
• Research.md is a mixed Markdown/YAML log. Each topic section has a small YAML block (topic, date, agent, word_count, verified) followed by the free-form interaction. The YAML lets you track what's been verified and by whom.
• Summary.md is ephemeral. Every bora write status run archives it. Don't invest in maintaining it by hand — that's the AI's job.
• ChapterProject.md drives the briefing. The status field (draft, in-progress, completed) and target_words feed the compiled status output. Keep them up to date.
• Chapter IDs are calculated, not counted. Deleting a chapter directory doesn't reset the counter — the next chapter always gets max(existing IDs) + 1. You won't accidentally reuse an ID.

Working across models

Bora is model-agnostic. It produces plain Markdown and YAML that any LLM can read.

Chat-only models (web Claude, ChatGPT, Gemini, etc.)

For dev projects, run bora dev context <project_path> --budget <N> and paste the output as your first message. The model gets the same complete briefing every time.

For write projects, run bora write status and paste the output. Ask the model to generate a Summary.md and save its response.

Agentic tools with file access (Claude Code, Cursor, Aider, etc.)

The model reads AGENTS.md and follows its instructions to discover the rest. For dev projects, the agent can run bora dev ticket set <project_path> …, bora dev lint <project_path>, and bora dev status <project_path> directly from its shell as work progresses.

Local models

The same flows work with local models. Smaller models (under ~14B parameters) may occasionally produce malformed YAML frontmatter in ticket or planning files. Run bora dev lint <project_path> after any model writes to a ticket file to catch these early.

Upgrading

From 0.3.x to 0.4.5

0.4.5 replaces the flat docs/ai/Project.md layout with hierarchical docs/ai/<Codebase>/<Target>/<Project>/ projects. Every dev command takes an explicit <project_path>. There is no active-project pointer and no bora dev project / bora dev decision command.

New work: bora dev init <project_path> [--tags …]. That creates a dated briefing, a dated Requirements file, per-project Status.md, and tickets/. There is no migration for existing flat trees — start a new hierarchical project alongside them if you still have the old files.

From 0.3.x to 0.3.5

0.3.5 added .bora/project.json to track the active project file. That pointer is unused in 0.4.5 hierarchical projects; <project_path> is authoritative.

Summary archives are named (YYYY-MM-DD) Summary.md instead of YYYY-MM-DD - Summary.md. Existing archives are not renamed.

From 0.2.x to 0.3.x

Bora 0.3.0 moved all commands under bora dev or bora write. The old top-level bora init prints a deprecation warning and exits.

1. Upgrade bora:

   pipx upgrade bora
   

2. Add a profile file. Run any bora dev command and choose dev at the prompt — bora writes the file automatically. Or create it manually:

   mkdir -p .bora
   cat > .bora/profile.json << 'EOF'
   {
     "version": "0.3.5",
     "profile": "dev",
     "initialized_at": "2026-01-01T00:00:00+00:00",
     "config": { "auto_archive": true, "research_log_mode": "full_interaction" }
   }
   EOF
   

3. All existing tickets, docs/ai/ files, and AGENTS.md are untouched. Commands that used to be bora ticket ... are now bora dev ticket ....

Contributing

The source is in bora/. Here's what each module does:

Tests live in tests/, one file per phase. Run them with:

python -m pytest tests/ -v

To manually smoke-test a dev project:

mkdir /tmp/test-bora && cd /tmp/test-bora && git init
bora dev init Acme/App
bora dev ticket new Acme/App "Test ticket" --priority high --no-edit
bora dev ticket set Acme/App 01 status in-progress
bora dev status Acme/App
bora dev lint Acme/App

To smoke-test a write project:

mkdir /tmp/test-write && cd /tmp/test-write
bora write init
bora write chapter "First Chapter"
bora write chapter "Second Chapter"
bora write status

License

MIT.