bd - Git-Native Issue Tracking

Core Principle

Use bd for all task and issue operations—it's git-native and agent-friendly.

Issues live where your code lives: in git. No external API, no network dependency, no vendor lock-in. Just files, tracked by git, queryable via SQLite, and optimized for AI agent workflows.

When to Use bd

Primary Use Cases

Task Tracking: Create, update, close issues for any work item
Issue Management: Organize bugs, features, chores with labels and priorities
Project Planning: Break down epics into subtasks with dependencies
Dependency Modeling: Express work relationships as a DAG
Sprint Planning: Assign work, set priorities, track progress
Bug Triage: Classify, assign, and track defects
AI Agent Workflows: Autonomous task execution with dependency awareness

When NOT to Use bd

Public issue tracking (use GitHub Issues for public visibility)
Customer-facing support tickets (use dedicated support systems)
Marketing/sales workflows (use CRM tools)
Document management (use proper DMS)

Quick Start

Installation Check

# Verify bd is installed
which bd
# Expected: /opt/homebrew/bin/bd (or similar)

bd --version
# Expected: bd version 0.28.0+

Initialize in Repository

# Initialize bd in current git repo
bd init

# Configure project-specific prefix (optional)
bd config set prefix myapp

# Verify setup
bd status

Basic Operations

# Create an issue
bd create "Add user authentication"

# Create with metadata
bd create "Fix login bug" \
  --type bug \
  --priority 1 \
  --assignee alice \
  --labels backend,security

# List issues
bd list
bd list --status open --assignee alice
bd list --label urgent --priority-min 0 --priority-max 1

# Update issue
bd update bd-1 --status in_progress
bd update bd-1 --priority 0
bd update bd-1 --assignee bob

# Add comment
bd comment bd-1 "Started investigating root cause"

# Close issue
bd close bd-1

# Show details
bd show bd-1
bd show bd-1 --json  # Structured output for parsing

Core Concepts

Issue Structure

Every issue has:

ID: Hash-based (e.g., bd-a3f8e9) or sequential (legacy)
Title: Brief summary (required)
Description: Detailed explanation (markdown)
Type: bug | feature | task | epic | chore
Status: open | in_progress | blocked | closed
Priority: 0-4 (0=highest, 4=lowest)
Assignee: Username or actor
Labels: Tags for organization
Dependencies: Relationships to other issues
Timestamps: created_at, updated_at, closed_at

Dual Persistence

bd uses TWO storage formats:

SQLite (.beads/*.db)
- Fast queries and filtering
- Full-text search
- Indexed relationships
- Ephemeral: rebuilds from JSONL
JSONL (.beads/beads.jsonl)
- Git-friendly line-by-line format
- Complete audit history
- Source of truth for sync
- Human-readable events

Key insight: SQLite is for speed, JSONL is for git. Changes are written to both automatically.

Dependency DAG

Issues form a Directed Acyclic Graph (no cycles allowed):

bd-1 (Design) → bd-2 (Backend) → bd-3 (Frontend) → bd-4 (Tests)

Dependency types:

blocks / blocked_by: Hard dependencies
discovered-from: Context tracking
parent / child: Hierarchical relationships

Why DAG?

Prevents deadlocks (no circular dependencies)
Enables critical path analysis
Clear work ordering
Parallel work identification

Issue Lifecycle

┌─────────────────────────────────────────┐
│                                         │
│  open ──> in_progress ──> closed       │
│    │            │                       │
│    └────> blocked                       │
│              │                          │
│              └──> in_progress           │
│                                         │
│  closed ──> reopen ──> open             │
│                                         │
└─────────────────────────────────────────┘

Common Workflows

Daily Developer Workflow

# Morning: Review work
git pull                              # Sync issues from team
bd ready                              # Show unblocked work
bd list --assignee $USER              # Your assigned issues

# During work: Update status
bd update bd-5 --status in_progress   # Start work
bd comment bd-5 "Implementing JWT"    # Progress notes

# Found a bug? Create it
bd create "Fix null pointer in auth" \
  --type bug \
  --priority 1 \
  --deps discovered-from:bd-5

# End of day: Sync
git add .beads/
git commit -m "Progress on bd-5"
git push

Epic Decomposition

# 1. Create epic
bd create "Redesign authentication" \
  --type epic \
  --priority 0

EPIC_ID="bd-1"  # Replace with actual ID

# 2. Create subtasks
bd create "Design auth schema" --parent $EPIC_ID --priority 0
bd create "Implement backend" --parent $EPIC_ID --deps bd-2
bd create "Update frontend" --parent $EPIC_ID --deps bd-3
bd create "Write tests" --parent $EPIC_ID --deps bd-3,bd-4

# 3. Visualize plan
bd dep tree $EPIC_ID --long

Sprint Planning

# 1. Review backlog
bd list --status open --no-assignee --sort priority

# 2. Assign sprint work
bd update bd-10,bd-11,bd-12 --assignee alice
bd label add sprint-5 bd-10,bd-11,bd-12

# 3. Set priorities
bd update bd-10,bd-11 --priority 0  # Critical

# 4. Review sprint
bd list --label sprint-5 --long

Bug Triage

# 1. Create bug with full context
bd create "Login fails with special chars" \
  --type bug \
  --priority 2 \
  --description "Steps: 1) Use p@ss\$word! 2) Try login 3) Fails" \
  --labels backend,security

# 2. Assign and track
bd update bd-20 --assignee backend-team --status in_progress

# 3. Add investigation notes
bd update bd-20 --notes "Root cause: encoding issue in auth/password.go"

# 4. Close when fixed
bd close bd-20

Dependency Management

Adding Dependencies

# Single dependency
bd dep add bd-1 blocks bd-2

# Multiple at creation
bd create "New feature" --deps bd-1,bd-2,bd-3

# Explicit types
bd create "Bug fix" --deps discovered-from:bd-5,blocks:bd-10

# Hierarchical
bd create "Subtask" --parent bd-1

Viewing Dependencies

# Dependency tree
bd dep tree bd-1

# All blocked issues
bd blocked

# All ready work (no blockers)
bd ready

# Export as graph
bd list --format dot | dot -Tpng -o issues.png
bd list --format digraph > issues.txt

Removing Dependencies

# Remove specific dependency
bd dep remove bd-1 blocks bd-2

Cycle Detection

# Find cycles (deadlocks)
bd dep cycles

# Output example:
# Cycle detected: bd-5 -> bd-7 -> bd-9 -> bd-5

Advanced Features

Daemon Mode

Long-running RPC server for hot database and multi-workspace:

# Start daemon (auto-started by default)
bd daemon start

# Check status
bd daemon status

# Stop daemon
bd daemon stop

# Force direct mode (bypass daemon)
bd --no-daemon list

Benefits:

<10ms create/update operations (vs ~100ms direct)
Hot SQLite cache
Multi-workspace support
Background sync
JSON-RPC protocol

Templates

# Create template
bd template create bug-report \
  --type bug \
  --priority 2 \
  --description "## Steps to Reproduce\n\n## Expected\n\n## Actual"

# Use template
bd create "New bug" --from-template bug-report

Labels

# Add labels
bd label add urgent bd-5
bd label add backend,security bd-5

# Remove labels
bd label remove urgent bd-5

# Query by labels
bd list --label backend         # AND: must have ALL
bd list --label-any urgent,high # OR: must have AT LEAST ONE

Search & Filtering

# Text search
bd search "authentication"

# Filter by multiple criteria
bd list \
  --status open,in_progress \
  --priority-min 0 --priority-max 1 \
  --assignee alice \
  --label backend \
  --created-after 2024-01-01 \
  --sort priority \
  --reverse

# Empty description (needs work)
bd list --empty-description

# No assignee (needs owner)
bd list --no-assignee --status open

Comments

# Add comment
bd comment bd-5 "Started implementation"
bd comment bd-5 --body "Multi-line comment"

# View comments
bd comments list bd-5

# Add inline (via update)
bd update bd-5 --notes "Design notes here"

Export/Import

# Export to JSONL
bd export > backup.jsonl
bd export --status closed --created-after 2024-01-01 > archive.jsonl

# Import from JSONL
bd import backup.jsonl

# Migrate between repos
bd migrate-issues --from ~/old-repo --to ~/new-repo

Multi-Repository

# Configure multiple repos
bd repo add backend ~/repos/backend --prefix api
bd repo add frontend ~/repos/frontend --prefix ui

# Create in specific repo
bd create "Add endpoint" --repo backend

# Auto-routing (based on cwd)
cd ~/repos/backend
bd create "Add endpoint"  # Automatically creates in backend repo

Git Integration

Automatic Sync

# bd auto-syncs with JSONL on operations
bd create "New issue"
# Writes to SQLite AND .beads/beads.jsonl

git add .beads/
git commit -m "Add issue"
git push

Git Hooks

# Install hooks for auto-sync
bd hooks install

# Hooks added:
# pre-commit: Validate issue references in commits
# post-commit: Auto-export to JSONL

Manual Sync

# Pull from remote
git pull  # Merges .beads/beads.jsonl

# Push to remote
git add .beads/
git commit -m "Update issues"
git push

# Or use bd sync
bd sync pull
bd sync push

Merge Conflicts

bd includes custom merge driver:

# .gitattributes (auto-configured by bd init)
.beads/beads.jsonl merge=beads

# Conflicts are resolved by:
# 1. Preserving both changes in JSONL
# 2. Rebuilding SQLite from merged JSONL
# 3. Detecting cycles if any

Visualization

bv (Interactive Graph)

# Open interactive visualization
bv .beads/beads.jsonl

# Features:
# - Colored nodes by status
# - Dependency edges with types
# - Filter by labels, status, assignee
# - Click nodes for details
# - Export as PNG/SVG

Graphviz

# Generate DOT format
bd list --format dot > issues.dot

# Render with Graphviz
dot -Tpng -o issues.png issues.dot
dot -Tsvg -o issues.svg issues.dot

# Customize layout
neato -Tpng -o issues-neato.png issues.dot  # Force-directed
circo -Tpng -o issues-circo.png issues.dot  # Circular

Text-Based Tree

# ASCII dependency tree
bd dep tree bd-1

# Long format (more details)
bd dep tree bd-1 --long

Health & Maintenance

Validation

# Full database health check
bd validate

# Check for issues:
# - Orphaned dependencies
# - Cycles in DAG
# - Corrupted records
# - JSONL/SQLite inconsistencies

Doctor

# Comprehensive health check
bd doctor

# Checks:
# - Installation integrity
# - Database schema
# - Config validity
# - Git integration
# - Daemon status

Repair

# Fix orphaned dependencies
bd repair-deps
bd repair-deps --dry-run  # Preview only

# Detect and clean test pollution
bd detect-pollution
bd detect-pollution --clean

# Find and merge duplicates
bd duplicates
bd duplicates --auto-merge

Cleanup

# Compact old closed issues (save space, keep git history)
bd compact --older-than 90d

# Delete closed issues from DB (keep in JSONL for sync)
bd cleanup --older-than 30d

# Purge deleted issues entirely (careful!)
bd delete bd-1 --purge  # Removes from JSONL too

AI Agent Workflows

Autonomous Task Loop

#!/bin/bash
# agent-work-loop.sh

while true; do
  # Get ready work
  READY=$(bd ready --json --assignee agent)

  if [ "$READY" = "[]" ]; then
    sleep 60
    continue
  fi

  # Pick highest priority
  ISSUE_ID=$(echo $READY | jq -r '.[0].id')

  # Mark in progress
  bd update $ISSUE_ID --status in_progress

  # Execute work (agent-specific)
  agent-execute $ISSUE_ID

  # Mark complete
  bd close $ISSUE_ID

  # Sync
  git add .beads/
  git commit -m "Completed $ISSUE_ID"
  git push
done

Context-Aware Creation

#!/bin/bash
# agent-discover-issues.sh

# Find TODOs in code
rg "TODO:" --json | jq -c '.[] | select(.type == "match")' | \
while read line; do
  FILE=$(echo $line | jq -r '.data.path.text')
  TEXT=$(echo $line | jq -r '.data.lines.text' | sed 's/.*TODO: //')

  bd create "$TEXT" \
    --type task \
    --priority 2 \
    --description "Found in $FILE" \
    --labels auto-generated
done

Dependency-Aware Planning

#!/bin/bash
# agent-plan-work.sh

# Build dependency graph
GRAPH=$(bd list --format digraph)

# Compute work order (topological sort)
echo "$GRAPH" | golang.org/x/tools/cmd/digraph allpaths | \
while read path; do
  echo "Execution order: $path"
done

JSON Output & Parsing

Structured Output

# Get JSON output (all commands support --json)
bd list --json
bd show bd-1 --json
bd ready --json
bd blocked --json

Example Parsing

# Get all high-priority open issues
bd list --status open --json | \
  jq '.[] | select(.priority <= 1) | {id, title, priority}'

# Find issues with no assignee
bd list --json | \
  jq '.[] | select(.assignee == null or .assignee == "") | .id'

# Count issues by status
bd list --json | \
  jq 'group_by(.status) | map({status: .[0].status, count: length})'

# Extract dependency graph
bd list --json | \
  jq '.[] | {id, blocks: [.dependencies[] | select(.type=="blocks") | .target_id]}'

JSONL Direct Access

Reading JSONL

For direct file access (e.g., in agent scripts):

# Read all events
cat .beads/beads.jsonl | jq -c '.'

# Filter by event type
cat .beads/beads.jsonl | jq -c 'select(.type == "create")'

# Get latest issue state
cat .beads/beads.jsonl | \
  jq -c 'select(.type == "create" or .type == "update")' | \
  jq -s 'group_by(.issue.id) | map(sort_by(.issue.updated_at) | last | .issue)'

JSONL Event Types

{"type":"create","issue":{...}}
{"type":"update","issue":{...}}
{"type":"close","id":"bd-1","closed_at":"...","actor":"..."}
{"type":"reopen","id":"bd-1","actor":"..."}
{"type":"delete","id":"bd-1","deleted_at":"...","actor":"..."}
{"type":"comment","comment":{...}}
{"type":"dep_add","dependency":{...}}
{"type":"dep_remove","dependency":{...}}

Configuration

Config Commands

# View current config
bd config list

# Set values
bd config set prefix myapp
bd config set default_priority 2
bd config set default_type task

# Git settings
bd config set git.auto_sync true
bd config set git.remote origin
bd config set git.sync_branch main

# Daemon settings
bd config set daemon.enabled true
bd config set daemon.sync_interval 300

Config File

Location: .beads/config.json (per-repo) or ~/.config/bd/config.json (global)

{
  "prefix": "myapp",
  "default_priority": 2,
  "default_type": "task",
  "daemon": {
    "enabled": true,
    "port": 9876,
    "sync_interval": 300
  },
  "git": {
    "auto_sync": true,
    "sync_branch": "main",
    "remote": "origin"
  },
  "repos": {
    "backend": {
      "path": "/Users/user/repos/backend",
      "prefix": "api",
      "auto_route": true
    },
    "frontend": {
      "path": "/Users/user/repos/frontend",
      "prefix": "ui",
      "auto_route": true
    }
  }
}

Statistics & Reporting

# Overall stats
bd stats

# Count issues
bd count
bd count --status open
bd count --label urgent

# Stale issues
bd stale --older-than 30d

# Export stats
bd list --json | \
  jq '{
    total: length,
    by_status: group_by(.status) | map({(.[0].status): length}) | add,
    by_type: group_by(.type) | map({(.[0].type): length}) | add,
    by_priority: group_by(.priority) | map({("\(.priority)"): length}) | add
  }'

Global Flags

--json              # JSON output for all commands
--db <path>         # Specify database path
--no-daemon         # Force direct mode (bypass daemon)
--no-db             # JSONL-only mode (no SQLite)
--no-auto-flush     # Disable auto-sync to JSONL
--no-auto-import    # Disable auto-import from JSONL
--sandbox           # Sandbox mode (no daemon, no auto-sync)
--quiet             # Suppress output
--verbose           # Debug output
--actor <name>      # Override actor name (audit trail)
--allow-stale       # Skip staleness check

Best Practices

1. Commit Issues with Code

git checkout -b feature/new-auth
bd create "Implement new auth" --type feature
# ... write code ...
bd update bd-1 --status in_progress
git add .
git commit -m "Implement new auth (bd-1)"
bd close bd-1
git add .beads/
git commit -m "Close bd-1"
git push

2. Use Dependencies Liberally

Model real work dependencies—it unlocks:

Critical path analysis
Parallel work identification
Automatic ready/blocked tracking

3. Label Consistently

Establish taxonomy early:

backend, frontend, infra
urgent, high, medium, low
sprint-N for sprint tracking
tech-debt, security, performance

4. Comment Frequently

Context decays over time. Comments preserve:

Why decisions were made
What was tried and failed
Links to resources
Status updates

5. Sync Daily

# Pull in morning
git pull

# Push at end of day
git add .beads/
git commit -m "Daily issue updates"
git push

6. Clean Up Regularly

# Weekly: review stale issues
bd stale --older-than 30d

# Monthly: compact old closed issues
bd compact --older-than 90d

# Quarterly: cleanup deleted issues
bd cleanup --older-than 90d

7. Validate Health

# Weekly: run validation
bd validate

# Fix issues immediately
bd repair-deps
bd dep cycles

Troubleshooting

Issues Not Appearing

# Check database exists
ls -la .beads/

# Validate JSONL
cat .beads/beads.jsonl | jq '.' > /dev/null

# Rebuild SQLite from JSONL
bd migrate

Daemon Not Starting

# Check daemon status
bd daemon status

# View logs
bd daemon logs

# Stop and restart
bd daemon stop
bd daemon start

Merge Conflicts

# After git pull with conflicts in .beads/beads.jsonl
git status

# bd's merge driver should handle automatically
# If not, rebuild from JSONL
rm .beads/*.db
bd migrate

# Validate
bd validate

Performance Issues

# Use daemon mode (default)
bd daemon status

# Compact old issues
bd compact --older-than 90d

# Check database size
du -sh .beads/

# Profile operations
bd --profile list

Related Tools

git: Version control (bd integrates natively)
bv: Interactive graph visualization for bd issues
sqlite3: Direct database queries if needed
jq: JSON processing for structured output
graphviz: Render dependency graphs (dot, neato, circo)
rg (ripgrep): Fast text search in issues
fd: Fast file search for bd files

References

For complete details, see:

Type Definitions: @bd-codebase/types/core.ts
Git-Native Principles: @bd-codebase/principles/git-native.md
DAG Dependencies: @bd-codebase/principles/dag-dependencies.md
Task Workflows: @bd-codebase/templates/task-workflow.md
Codebase README: @bd-codebase/README.md

Quick Reference

See assets/cheatsheet.md for one-page reference.

Remember: bd is designed to work like git because it is git. Think of issues as files, operations as commits, and sync as push/pull. It's that simple.