Give your coding agent a memory upgrade

🎉 v0.20.1: Multi-Worker Support Unlocked! 🎉

Hash-based IDs eliminate merge conflicts and collision issues!

Previous versions used sequential IDs (bd-1, bd-2, bd-3...) which caused frequent collisions when multiple agents or branches created issues concurrently. Version 0.20.1 switches to hash-based IDs (bd-a1b2, bd-f14c, bd-3e7a...) that are collision-resistant and merge-friendly.

What's new: ✅ Multi-clone, multi-branch, multi-agent workflows now work reliably
What changed: Issue IDs are now short hashes instead of sequential numbers
Migration: Run bd migrate to upgrade existing databases (optional - old DBs still work)

Hash IDs use progressive length scaling (4/5/6 characters) with birthday paradox math to keep collisions extremely rare while maintaining human readability. See "Hash-Based Issue IDs" section below for details.

⚠️ Alpha Status: This project is in active development. The core features work well, but expect API changes before 1.0. Use for development/internal projects first.

Beads is a lightweight memory system for coding agents, using a graph-based issue tracker. Four kinds of dependencies work to chain your issues together like beads, making them easy for agents to follow for long distances, and reliably perform complex task streams in the right order.

Drop Beads into any project where you're using a coding agent, and you'll enjoy an instant upgrade in organization, focus, and your agent's ability to handle long-horizon tasks over multiple compaction sessions. Your agents will use issue tracking with proper epics, rather than creating a swamp of rotten half-implemented markdown plans.

Instant start:

curl -fsSL https://raw.githubusercontent.com/steveyegge/beads/main/scripts/install.sh | bash

Then tell your coding agent to start using the bd tool instead of markdown for all new work, somewhere in your AGENTS.md or CLAUDE.md. That's all there is to it!

You don't use Beads directly as a human. Your coding agent will file and manage issues on your behalf. They'll file things they notice automatically, and you can ask them at any time to add or update issues for you.

Beads gives agents unprecedented long-term planning capability, solving their amnesia when dealing with complex nested plans. They can trivially query the ready work, orient themselves, and land on their feet as soon as they boot up.

Agents using Beads will no longer silently pass over problems they notice due to lack of context space -- instead, they will automatically file issues for newly-discovered work as they go. No more lost work, ever.

Beads issues are backed by git, but through a clever design it manages to act like a managed, centrally hosted SQL database shared by all of the agents working on a project (repo), even across machines.

Beads even improves work auditability. The issue tracker has a sophisticated audit trail, which agents can use to reconstruct complex operations that may have spanned multiple sessions.

Agents report that they enjoy working with Beads, and they will use it spontaneously for both recording new work and reasoning about your project in novel ways. Whether you are a human or an AI, Beads lets you have more fun and less stress with agentic coding.

• ✨ Zero setup - bd init creates project-local database (and your agent will do it)
• 🔗 Dependency tracking - Four dependency types (blocks, related, parent-child, discovered-from)
• 📋 Ready work detection - Automatically finds issues with no open blockers
• 🤖 Agent-friendly - --json flags for programmatic integration
• 📦 Git-versioned - JSONL records stored in git, synced across machines
• 🌍 Distributed by design - Agents on multiple machines share one logical database via git
• 🔐 Protected branch support - Works with GitHub/GitLab protected branches via separate sync branch
• 🏗️ Extensible - Add your own tables to the SQLite database
• 🔍 Multi-project isolation - Each project gets its own database, auto-discovered by directory
• 🌲 Dependency trees - Visualize full dependency graphs
• 🎨 Beautiful CLI - Colored output for humans, JSON for bots
• 💾 Full audit trail - Every change is logged
• ⚡ High performance - Batch operations for bulk imports (1000 issues in ~950ms)
• 🗜️ Memory decay - Semantic compaction gracefully reduces old closed issues

npm (Node.js environments, Claude Code for Web):

npm install -g @beads/bd

Quick install (all platforms):

curl -fsSL https://raw.githubusercontent.com/steveyegge/beads/main/scripts/install.sh | bash

Homebrew (macOS/Linux):

brew tap steveyegge/beads
brew install bd

Other platforms and methods: See INSTALLING.md for Windows, Arch Linux, and manual installation.

IDE Integration: See INSTALLING.md for Claude Code plugin and MCP server setup.

Claude Code for Web: See npm-package/CLAUDE_CODE_WEB.md for SessionStart hook setup.

Beads is designed for AI coding agents to use on your behalf. Setup takes 30 seconds:

You run this once (humans only):

# In your project root:
bd init

# For protected branches (GitHub/GitLab):
bd init --branch beads-metadata

# bd will:
# - Create .beads/ directory with database
# - Import existing issues from git (if any)
# - Prompt to install git hooks (recommended: say yes)
# - Auto-start daemon for sync

# Then tell your agent about bd:
echo "BEFORE ANYTHING ELSE: run 'bd onboard' and follow the instructions" >> AGENTS.md

Protected branches? If your main branch is protected, use bd init --branch beads-metadata to commit issue updates to a separate branch. See docs/PROTECTED_BRANCHES.md for details.

Your agent does the rest: Next time your agent starts, it will:

1. Run bd onboard and receive integration instructions
2. Add bd workflow documentation to AGENTS.md
3. Update CLAUDE.md with a note (if present)
4. Remove the bootstrap instruction

For agents setting up repos: Use bd init --quiet for non-interactive setup (auto-installs git hooks, no prompts).

For new repo clones: Run bd init (or bd init --quiet for agents) to import existing issues from .beads/issues.jsonl automatically.

Most tasks will be created and managed by agents during conversations. You can check on things with:

bd list                  # See what's being tracked
bd show <issue-id>       # Review a specific issue
bd ready                 # See what's ready to work on
bd dep tree <issue-id>   # Visualize dependencies

Run the interactive guide to learn the full workflow:

bd quickstart

Quick reference for agent workflows:

# Find ready work
bd ready --json | jq '.[0]'

# Create issues during work
bd create "Discovered bug" -t bug -p 0 --json

# Link discovered work back to parent
bd dep add <new-id> <parent-id> --type discovered-from

# Update status
bd update <issue-id> --status in_progress --json

# Complete work
bd close <issue-id> --reason "Implemented" --json

Here's the crazy part: bd acts like a centralized database, but it's actually distributed via git.

When you install bd on any machine with your project repo, you get:

• ✅ Full query capabilities (dependencies, ready work, etc.)
• ✅ Fast local operations (<100ms via SQLite)
• ✅ Shared state across all machines (via git)
• ✅ No server, no daemon required, no configuration
• ✅ AI-assisted merge conflict resolution

How it works: Each machine has a local SQLite cache (.beads/*.db, gitignored). Source of truth is JSONL (.beads/issues.jsonl, committed to git). Auto-sync keeps them in sync: SQLite → JSONL after CRUD operations (5-second debounce), JSONL → SQLite when JSONL is newer (e.g., after git pull).

The result: Agents on your laptop, your desktop, and your coworker's machine all query and update what feels like a single shared database, but it's really just git doing what git does best - syncing text files across machines.

No PostgreSQL instance. No MySQL server. No hosted service. Just install bd, clone the repo, and you're connected to the "database."

bd automatically syncs your local database with git:

Making changes (auto-export):

bd create "Fix bug" -p 1
bd update bd-a1b2 --status in_progress
# bd automatically exports to .beads/issues.jsonl after 5 seconds

git add .beads/issues.jsonl
git commit -m "Working on bd-a1b2"
git push

Pulling changes (auto-import):

git pull
# bd automatically detects JSONL is newer and imports on next command

bd ready  # Fresh data from git!
bd list   # Shows issues from other machines

Manual sync (optional):

bd sync  # Immediately flush pending changes and import latest JSONL

For zero-lag sync, install the git hooks:

cd examples/git-hooks && ./install.sh

This adds:

• pre-commit - Immediate flush before commit (no 5-second wait)
• post-merge - Guaranteed import after git pull or git merge

Disable auto-sync if needed:

bd --no-auto-flush create "Issue"   # Skip auto-export
bd --no-auto-import list            # Skip auto-import check

Version 0.20.1 introduces collision-resistant hash-based IDs to enable reliable multi-worker and multi-branch workflows.

Issue IDs now use short hexadecimal hashes instead of sequential numbers:

• Before (v0.20.0): bd-1, bd-2, bd-152 (sequential numbers)
• After (v0.20.1): bd-a1b2, bd-f14c, bd-3e7a (4-6 character hashes)

Hash IDs use progressive length scaling based on database size:

• 0-500 issues: 4-character hashes (e.g., bd-a1b2)
• 500-1,500 issues: 5-character hashes (e.g., bd-f14c3)
• 1,500-10,000 issues: 6-character hashes (e.g., bd-3e7a5b)

The problem with sequential IDs: When multiple agents or branches create issues concurrently, sequential IDs collide:

• Agent A creates bd-10 on branch feature-auth
• Agent B creates bd-10 on branch feature-payments
• Git merge creates duplicate IDs → collision resolution required

How hash IDs solve this: Hash IDs are generated from random data, making concurrent creation collision-free:

• Agent A creates bd-a1b2 (hash of random UUID)
• Agent B creates bd-f14c (different hash, different UUID)
• Git merge succeeds cleanly → no collision resolution needed

Hash IDs use birthday paradox probability to determine length:

Our thresholds are conservative: We switch from 4→5 chars at 500 issues (way before the 1% collision point at ~1,217) and from 5→6 chars at 1,500 issues.

Progressive extension on collision: If a hash collision does occur, bd automatically extends the hash to 7 or 8 characters instead of remapping to a new ID.

Existing databases continue to work - no forced migration. Run bd migrate when ready:

# Inspect migration plan (for AI agents)
bd migrate --inspect --json

# Check schema and config state
bd info --schema --json

# Preview migration
bd migrate --dry-run

# Migrate database schema (removes obsolete issue_counters table)
bd migrate

# Show current database info
bd info

AI-supervised migrations: The --inspect flag provides migration plan analysis for AI agents. The system verifies data integrity invariants (required config keys, foreign key constraints, issue counts) before committing migrations.

Note: Hash IDs require schema version 9+. The bd migrate command detects old schemas and upgrades automatically.

Hash IDs support hierarchical children for natural work breakdown structures. Child IDs use dot notation:

bd-a3f8e9      [epic] Auth System
bd-a3f8e9.1    [task] Design login UI
bd-a3f8e9.2    [task] Backend validation
bd-a3f8e9.3    [epic] Password Reset
bd-a3f8e9.3.1  [task] Email templates
bd-a3f8e9.3.2  [task] Reset flow tests

Benefits:

• Collision-free: Parent hash ensures unique namespace
• Human-readable: Clear parent-child relationships
• Flexible depth: Up to 3 levels of nesting
• No coordination needed: Each epic owns its child ID space

Common patterns:

• 1 level: Epic → tasks (most projects)
• 2 levels: Epic → features → tasks (large projects)
• 3 levels: Epic → features → stories → tasks (complex projects)

Example workflow:

# Create parent epic (generates hash ID automatically)
bd create "Auth System" -t epic -p 1
# Returns: bd-a3f8e9

# Create child tasks
bd create "Design login UI" -p 1       # Auto-assigned: bd-a3f8e9.1
bd create "Backend validation" -p 1    # Auto-assigned: bd-a3f8e9.2

# Create nested epic with its own children
bd create "Password Reset" -t epic -p 1  # Auto-assigned: bd-a3f8e9.3
bd create "Email templates" -p 1          # Auto-assigned: bd-a3f8e9.3.1

Note: Child IDs are automatically assigned sequentially within each parent's namespace. No need to specify parent manually - bd tracks context from git branch/working directory.

Check installation health: bd doctor validates your .beads/ setup, database version, ID format, and CLI version. Provides actionable fixes for any issues found.

bd create "Fix bug" -d "Description" -p 1 -t bug
bd create "Add feature" --description "Long description" --priority 2 --type feature
bd create "Task" -l "backend,urgent" --assignee alice

# Get JSON output for programmatic use
bd create "Fix bug" -d "Description" --json

# Create from templates (built-in: epic, bug, feature)
bd create --from-template epic "Q4 Platform Improvements"
bd create --from-template bug "Auth token validation fails"
bd create --from-template feature "Add OAuth support"

# Override template defaults
bd create --from-template bug "Critical issue" -p 0  # Override priority

# Create multiple issues from a markdown file
bd create -f feature-plan.md

Options:

• -f, --file - Create multiple issues from markdown file
• --from-template - Use template (epic, bug, feature, or custom)
• -d, --description - Issue description
• -p, --priority - Priority (0-4, 0=highest, default=2)
• -t, --type - Type (bug|feature|task|epic|chore, default=task)
• -a, --assignee - Assign to user
• -l, --labels - Comma-separated labels
• --id - Explicit issue ID (e.g., worker1-100 for ID space partitioning)
• --json - Output in JSON format

See bd template list for available templates and bd help template for managing custom templates.

bd info                                    # Show database path and daemon status
bd show bd-a1b2                            # Show full details
bd list                                    # List all issues
bd list --status open                      # Filter by status
bd list --priority 1                       # Filter by priority
bd list --assignee alice                   # Filter by assignee
bd list --label=backend,urgent             # Filter by labels (AND)
bd list --label-any=frontend,backend       # Filter by labels (OR)

# Advanced filters
bd list --title-contains "auth"            # Search title
bd list --desc-contains "implement"        # Search description
bd list --notes-contains "TODO"            # Search notes
bd list --id bd-123,bd-456                 # Specific IDs (comma-separated)

# Date range filters (YYYY-MM-DD or RFC3339)
bd list --created-after 2024-01-01         # Created after date
bd list --created-before 2024-12-31        # Created before date
bd list --updated-after 2024-06-01         # Updated after date
bd list --updated-before 2024-12-31        # Updated before date
bd list --closed-after 2024-01-01          # Closed after date
bd list --closed-before 2024-12-31         # Closed before date

# Empty/null checks
bd list --empty-description                # Issues with no description
bd list --no-assignee                      # Unassigned issues
bd list --no-labels                        # Issues with no labels

# Priority ranges
bd list --priority-min 0 --priority-max 1  # P0 and P1 only
bd list --priority-min 2                   # P2 and below

# Combine multiple filters
bd list --status open --priority 1 --label-any urgent,critical --no-assignee

# JSON output for agents
bd info --json
bd list --json
bd show bd-a1b2 --json

bd update bd-a1b2 --status in_progress
bd update bd-a1b2 --priority 2
bd update bd-a1b2 --assignee bob
bd close bd-a1b2 --reason "Completed"
bd close bd-a1b2 bd-f14c bd-3e7a   # Close multiple

# JSON output
bd update bd-a1b2 --status in_progress --json

# Add dependency (bd-f14c depends on bd-a1b2)
bd dep add bd-f14c bd-a1b2
bd dep add bd-3e7a bd-a1b2 --type blocks

# Remove dependency
bd dep remove bd-f14c bd-a1b2

# Show dependency tree
bd dep tree bd-f14c

# Detect cycles
bd dep cycles

• blocks: Hard blocker (default) - issue cannot start until blocker is resolved
• related: Soft relationship - issues are connected but not blocking
• parent-child: Hierarchical relationship (child depends on parent)
• discovered-from: Issue discovered during work on another issue (automatically inherits parent's source_repo)

Only blocks dependencies affect ready work detection.

Note: Issues created with discovered-from dependencies automatically inherit the parent's source_repo field, ensuring discovered work stays in the same repository as the parent task.

# Show ready work (no blockers)
bd ready
bd ready --limit 20
bd ready --priority 1
bd ready --assignee alice

# Sort policies (hybrid is default)
bd ready --sort priority    # Strict priority order (P0, P1, P2, P3)
bd ready --sort oldest      # Oldest issues first (backlog clearing)
bd ready --sort hybrid      # Recent by priority, old by age (default)

# Show blocked issues
bd blocked

# Statistics
bd stats

# JSON output for agents
bd ready --json

Add flexible metadata to issues for filtering and organization:

# Add labels during creation
bd create "Fix auth bug" -t bug -p 1 -l auth,backend,urgent

# Add/remove labels
bd label add bd-a1b2 security
bd label remove bd-a1b2 urgent

# List labels
bd label list bd-a1b2            # Labels on one issue
bd label list-all                # All labels with counts

# Filter by labels
bd list --label backend,auth     # AND: must have ALL labels
bd list --label-any frontend,ui  # OR: must have AT LEAST ONE

See LABELS.md for complete label documentation and best practices.

# Single issue deletion (preview mode)
bd delete bd-a1b2

# Force single deletion
bd delete bd-a1b2 --force

# Batch deletion
bd delete bd-a1b2 bd-f14c bd-3e7a --force

# Delete from file (one ID per line)
bd delete --from-file deletions.txt --force

# Cascade deletion (recursively delete dependents)
bd delete bd-a1b2 --cascade --force

The delete operation removes all dependency links, updates text references to [deleted:ID], and removes the issue from database and JSONL.

Manage per-project configuration for external integrations:

# Set configuration
bd config set jira.url "https://company.atlassian.net"
bd config set jira.project "PROJ"

# Get configuration
bd config get jira.url

# List all configuration
bd config list --json

# Unset configuration
bd config unset jira.url

See CONFIG.md for complete configuration documentation.

Beads uses AI to compress old closed issues, keeping databases lightweight as they age:

bd compact --dry-run --all  # Preview candidates
bd compact --days 90        # Compact closed issues older than 90 days

This is agentic memory decay - your database naturally forgets fine-grained details while preserving essential context.

# Export to JSONL (automatic by default)
bd export -o issues.jsonl

# Import from JSONL (automatic when JSONL is newer)
bd import -i issues.jsonl

# Manual sync
bd sync

Note: Auto-sync is enabled by default. Manual export/import is rarely needed.

bd runs a background daemon per workspace for auto-sync and RPC operations. Use bd daemons to manage multiple daemons:

# List all running daemons
bd daemons list

# Check health (version mismatches, stale sockets)
bd daemons health

# Stop a specific daemon
bd daemons stop /path/to/workspace
bd daemons stop 12345  # By PID

# View daemon logs
bd daemons logs /path/to/workspace -n 100
bd daemons logs 12345 -f  # Follow mode

# Stop all daemons
bd daemons killall
bd daemons killall --force  # Force kill if graceful fails

Common use cases:

• After upgrading bd: Run bd daemons health to check for version mismatches, then bd daemons killall to restart all daemons with the new version
• Debugging: Use bd daemons logs <workspace> to view daemon logs
• Cleanup: bd daemons list auto-removes stale sockets

See commands/daemons.md for complete documentation.

Check out the examples/ directory for:

• Python agent - Full agent implementation in Python
• Bash agent - Shell script agent example
• Git hooks - Automatic export/import on git operations
• Branch merge workflow - Handle ID collisions when merging branches
• Claude Desktop MCP - MCP server for Claude Desktop
• Claude Code Plugin - One-command installation with slash commands

For advanced usage, see:

• ADVANCED.md - Prefix renaming, merging duplicates, daemon configuration
• CONFIG.md - Configuration system for integrations
• EXTENDING.md - Database extension patterns
• ADVANCED.md - JSONL format and merge strategies

• README.md - You are here! Core features and quick start
• INSTALLING.md - Complete installation guide for all platforms
• QUICKSTART.md - Interactive tutorial (bd quickstart)
• FAQ.md - Frequently asked questions
• TROUBLESHOOTING.md - Common issues and solutions
• ADVANCED.md - Advanced features and use cases
• LABELS.md - Complete label system guide
• CONFIG.md - Configuration system
• EXTENDING.md - Database extension patterns
• ADVANCED.md - JSONL format analysis
• PLUGIN.md - Claude Code plugin documentation
• CONTRIBUTING.md - Contribution guidelines
• SECURITY.md - Security policy

• Beadster - Native macOS app for viewing and managing bd issues across multiple projects. Features a compact, always-on-top window for quick reference during development. Built by @podviaznikov.

Have you built something cool with bd? Open an issue to get it featured here!

# Run tests
go test ./...

# Build
go build -o bd ./cmd/bd

# Run
./bd create "Test issue"

# Bump version
./scripts/bump-version.sh 0.9.3           # Update all versions, show diff
./scripts/bump-version.sh 0.9.3 --commit  # Update and auto-commit

See scripts/README.md for more development scripts.

MIT

Built with ❤️ by developers who love tracking dependencies and finding ready work.

Inspired by the need for a simpler, dependency-aware issue tracker.