CLI Commands

All dtoolkit CLIs follow the same conventions: commander for argument parsing, picocolors for output, and @clack/prompts for interactive wizards. Install any CLI globally with npm install -g @dtoolkit/<package> or run directly via npx.

Tip

Every CLI supports --help for detailed usage and --version to check the installed version.

dbrain

Persistent memory server. Manages brain initialization, server lifecycle, federation, and API keys.

Commands

Command	Description
dbrain init [path]	Initialize a new brain (database, config, identity)
dbrain start [path]	Start the API server (REST + MCP on :7878) and dashboard (:7879)
dbrain connect <client> [url]	Connect a client CLI to a running brain
dbrain status [path]	Check brain health and status
dbrain compact	Compact the brain (deduplicate facts, recalculate tiers)
dbrain configure [path]	Reconfigure the brain interactively
dbrain link <url>	Link to a shared brain (federation)
dbrain unlink <name>	Remove a connection to a shared brain
dbrain connections	List connected brains with health status
dbrain keys <action>	Manage API keys for shared brains (create, list, revoke)

init

Interactive wizard that creates a new brain. Sets up the SQLite database, config file, identity documents, and initial entities.

terminal

dbrain init
dbrain init /custom/data/path
dbrain init --non-interactive

Flag	Description
[path]	Data path (default: ~/.dbrain)
--non-interactive	Skip prompts, use environment variables (for Docker/CI)

Environment variables for non-interactive mode:

Variable	Default	Purpose
DBRAIN_DATA	~/.dbrain	Data path
DBRAIN_PORT	7878	API port
DBRAIN_HOST	0.0.0.0	Bind address
DBRAIN_TOKEN	Auto-generated	Access token
DBRAIN_AGENT_NAME	dBrain	AI identity name
DBRAIN_OWNER_NAME	Human	Owner name
DBRAIN_BRAIN_TYPE	personal	personal or shared

connect

Configures a client CLI to connect to a running brain. Fetches connection details from the brain’s /connect endpoint and writes the client’s native config files.

terminal

dbrain connect claude http://localhost:7878
dbrain connect gemini http://your-server:7878 --token sk-dbr_abc123

Flag	Description
<client>	Target CLI: claude, codex, gemini, opencode
[url]	Brain URL (prompted if omitted)
--token <token>	Access token

compact

Runs brain maintenance: deduplicates cold facts, recalculates memory tiers, and optionally extracts entities from unprocessed conversations.

terminal

dbrain compact
dbrain compact --process --dry-run
dbrain compact --dedup --threshold 0.9

Flag	Description
--path <path>	Data path
--process	Extract entities and facts from unprocessed conversations via LLM
--dedup	Deduplicate cold facts only
--tiers	Recalculate tiers only
--dry-run	Preview changes without writing
--threshold <n>	Similarity threshold, 0-1 (default: 0.85)
--limit <n>	Max facts per batch (default: 1000)

link / unlink / connections

Manage federation between personal and shared brains.

terminal

dbrain link http://team-brain:7878 --token sk-dbr_team123
dbrain unlink team-brain
dbrain connections

Flag	Description
--token <token>	Access token for the shared brain
--name <name>	Connection name (defaults to remote brain name)
--path <path>	Local brain data path

keys

Manage per-user API keys for shared brains.

terminal

dbrain keys create --user-id dev1 --user-name "Developer One"
dbrain keys list
dbrain keys revoke --key-id abc123

Flag	Description
<action>	create, list, or revoke
--url <url>	Brain URL (or DBRAIN_URL env)
--token <token>	Admin token (or DBRAIN_TOKEN env)
--user-id <id>	User ID (for create)
--user-name <name>	Display name (for create)
--permissions <perms>	read, write, or read+write (default: read+write)
--key-id <id>	Key ID (for revoke)

---

dcontext

Hooks for AI coding CLIs. Injects identity and project context at session start, saves transcripts before compaction.

Commands

Command	Description
dcontext init	Initialize dcontext for the current directory
dcontext install <target>	Install hooks for a target CLI
dcontext uninstall <target>	Remove hooks for a target CLI
dcontext status	Show configuration, context, and stats
dcontext config	Modify dcontext configuration interactively
dcontext explore [path]	Preview the context that would be injected for a directory
dcontext sync	Sync project code graph
dcontext hook <event>	Handle hook events (internal, called by CLI hooks)

init

Interactive wizard that connects dcontext to dbrain (and optionally dwork and dops), maps the current directory to an entity, and saves configuration.

terminal

dcontext init

The wizard will prompt to connect to an existing dbrain or start a local one, then optionally connect to dwork and dops servers.

install / uninstall

Register or remove dcontext hooks in a target CLI’s configuration. After installation, the target CLI will automatically receive identity and project context at session start.

terminal

dcontext install claude
dcontext install gemini
dcontext install opencode
dcontext uninstall claude

Flag	Description
<target>	Target CLI: claude, gemini, opencode

explore

Preview the full context block that dcontext would inject for a given directory, without actually modifying any files. Useful for debugging context injection.

terminal

dcontext explore
dcontext explore /path/to/project

status

Displays current configuration including connected services (dbrain, dwork, dops), installed targets, entity mappings, and injection statistics.

terminal

dcontext status

---

dwork

AI-native, Markdown-driven project manager. Tasks live in BACKLOG.md, documentation in numbered files. SQLite FTS5 provides search.

Commands

Command	Description
dwork init [path]	Initialize dwork (config, database, optional first project)
dwork start [path]	Start the server (REST + MCP on :7881, dashboard on :7882)
dwork status [path]	Check dwork health and status
dwork sync <project>	Sync project docs from source code via dproxy
dwork connect [client] [url]	Connect a client CLI to a running dwork server
dwork configure [path]	Reconfigure dwork interactively
dwork keys <action>	Manage API keys (create, list, revoke)

init

Interactive wizard that creates the config file, SQLite database, and optionally scaffolds a first project with the four mandatory Markdown files (README.md, TECH.md, ROADMAP.md, BACKLOG.md).

terminal

dwork init
dwork init /custom/data/path
dwork init --non-interactive

Flag	Description
[path]	Data path (default: ~/.dwork)
--non-interactive	Skip prompts (for Docker/CI)

sync

Syncs project documentation from the source code directory by analyzing it via dproxy. Requires dproxy to be configured in config.json.

terminal

dwork sync my-project
dwork sync my-project --path /custom/data

Flag	Description
<project>	Project slug
--path <path>	Data path

connect

Configures a client CLI to use a running dwork server via MCP.

terminal

dwork connect claude
dwork connect gemini http://team-server:7881 --token sk-dwk_abc123

Flag	Description
[client]	Target CLI: claude, codex, gemini, opencode
[url]	dwork URL (prompted if omitted)
--token <token>	Access token

keys

Manage API keys for the dwork server.

terminal

dwork keys create --user-id dev1 --user-name "Alice"
dwork keys list
dwork keys revoke --key-id abc123

Flag	Description
<action>	create, list, or revoke
--url <url>	dwork URL (or DWORK_URL env)
--token <token>	Admin token (or DWORK_TOKEN env)
--user-id <id>	User ID (for create)
--user-name <name>	Display name (for create)
--permissions <perms>	read, write, or read+write (default: read+write)
--key-id <id>	Key ID (for revoke)

---

dproxy

Universal CLI adapter for invoking AI models via local CLIs. Supports four providers with persistent memory, conversation history, and prompt templates.

Commands

Command	Description
dproxy init	Interactive setup wizard (required before first use)
dproxy ask [prompt...]	Send a single-shot prompt to an AI model
dproxy chat	Start an interactive conversation (REPL)
dproxy serve	Start the dproxy HTTP API server
dproxy history <subcommand>	Manage query history (list, show, search, clear)
dproxy memory <subcommand>	Manage persistent memory snippets (set, get, list, search, delete)
dproxy template <subcommand>	Manage YAML prompt templates (list, show, add, run, delete)
dproxy config [set|get]	View or modify configuration

Tip

You can skip the ask subcommand entirely. Running dproxy "your question" is a shorthand for dproxy ask "your question".

ask

Send a prompt and receive a response. Supports file attachments, streaming, session continuation, and context injection.

terminal

dproxy ask "Explain this error" -p claude
dproxy ask "Review this code" --file src/main.ts --stream
echo "data here" | dproxy ask "Summarize this"
dproxy ask "Continue from before" --continue

Flag	Description
-p, --provider <name>	Provider: claude, codex, gemini, opencode
-m, --model <model>	Model override
--stream	Stream response text as it arrives
--system-prompt <text>	System prompt override
-f, --file <paths...>	Attach one or more files to the prompt
-o, --output-format <fmt>	Output format: text, json, stream-json
--max-turns <n>	Max agent turns
--max-budget-usd <n>	Max budget in USD
--no-memory	Skip memory injection
--memory <keys>	Inject only specific memory keys (comma-separated)
--no-life	Skip life/PARA context injection
--no-history	Do not save to history
--raw	Print raw JSON response
--token-footer	Append token usage footer
--max-session-tokens <n>	Reset session if context exceeds this token count
-c, --continue	Continue last conversation
-r, --resume <id>	Resume a specific session

chat

Interactive REPL for multi-turn conversations. Shares the same context injection pipeline as ask.

terminal

dproxy chat
dproxy chat -p gemini --continue
dproxy chat --resume abc123 -m claude-sonnet-4-20250514

Flag	Description
-p, --provider <name>	Provider: claude, codex, gemini, opencode
-m, --model <model>	Model override
-c, --continue	Continue last conversation
-r, --resume <id>	Resume a specific session
--max-turns <n>	Max agent turns per message
--no-memory	Skip memory injection
--no-life	Skip life/PARA context injection

serve

Start a Fastify HTTP API server that provides full CLI parity via REST endpoints with Bearer token auth and SSE streaming.

terminal

dproxy serve
dproxy serve --port 8080 --host 0.0.0.0

Flag	Description
--port <port>	Port to listen on (default from config)
--host <host>	Host to bind to

history

Manage the append-only JSONL query history.

terminal

dproxy history list
dproxy history list --limit 50
dproxy history show abc12345
dproxy history search "react hooks"
dproxy history clear

Subcommand	Description
list	List recent queries (--limit <n>, default 20)
show <id>	Show full prompt, response, cost, and duration
search <query>	Search history entries
clear	Delete all history

memory

CRUD for named memory snippets stored as Markdown files. Memory is automatically injected into prompts unless --no-memory is used.

terminal

dproxy memory set project-conventions "Always use TypeScript strict mode"
dproxy memory get project-conventions
dproxy memory list
dproxy memory search "typescript"
dproxy memory delete project-conventions

Subcommand	Description
set <key> <value...>	Create or update a memory entry
get <key>	Read a memory entry
list	List all memory keys
search <query>	Search memory contents
delete <key>	Remove a memory entry

template

CRUD for YAML prompt templates with {{variable}} interpolation.

terminal

dproxy template list
dproxy template show code-review
dproxy template add code-review --file review.yaml
dproxy template run code-review --var file=src/main.ts
dproxy template delete code-review

Subcommand	Description
list	List all templates
show <name>	Show template details
add <name>	Create a template (from file or stdin)
run <name>	Execute a template with variable substitution
delete <name>	Remove a template

---

dops

Agent observability. Ingests telemetry from AI CLI transcripts and provides usage statistics, cost breakdowns, and error tracking.

Commands

Command	Description
dops init [path]	Initialize dops (config and database)
dops start [path]	Start the server (REST + MCP on :7883, dashboard on :7884)
dops status [path]	Check dops health and status
dops ingest	Scan local CLI transcripts and push telemetry to the server
dops stats	Show agent usage statistics
dops costs	Show estimated costs by model

init

Creates the config file and SQLite database for storing telemetry data.

terminal

dops init
dops init /custom/data/path
dops init --non-interactive

Flag	Description
[path]	Data path (default: ~/.dops)
--non-interactive	Skip prompts (for Docker/CI)

ingest

Scans local transcript directories for Claude Code, Codex, Gemini, and OpenCode sessions, then pushes the extracted telemetry to a running dops server.

terminal

dops ingest --url http://localhost:7883 --token sk-dops_abc123
dops ingest --url http://localhost:7883 --token sk-dops_abc123 --days 7 --source claude gemini
dops ingest --url http://localhost:7883 --token sk-dops_abc123 --dry-run

Flag	Description
--url <url>	dops server URL (required)
--token <token>	dops access token (required)
--days <n>	Scan last N days (default: 30)
--dry-run	Scan and report without pushing data
--source <names...>	Filter sources: claude, codex, gemini, opencode

stats

Display aggregated usage statistics including tokens, tool calls, and errors.

terminal

dops stats --url http://localhost:7883 --token sk-dops_abc123
dops stats --url http://localhost:7883 --token sk-dops_abc123 --days 7 --source claude

Flag	Description
--url <url>	dops server URL (required)
--token <token>	dops access token (required)
--days <n>	Last N days (default: 30)
--source <name>	Filter by source: claude, codex, gemini, opencode

costs

Show estimated costs broken down by model, calculated from token usage and published pricing.

terminal

dops costs --url http://localhost:7883 --token sk-dops_abc123
dops costs --url http://localhost:7883 --token sk-dops_abc123 --days 7

Flag	Description
--url <url>	dops server URL (required)
--token <token>	dops access token (required)
--days <n>	Last N days (default: 30)

---

Quick reference

A summary of all CLI entry points and the default ports for server products.

Product	Install	Default port(s)
dbrain	npm install -g @dtoolkit/dbrain	:7878 (API + MCP), :7879 (Dashboard)
dcontext	npm install -g @dtoolkit/dcontext	N/A (hooks only)
dwork	npm install -g @dtoolkit/dwork	:7881 (API + MCP), :7882 (Dashboard)
dproxy	npm install -g @dtoolkit/dproxy	:7880 (HTTP API via serve)
dops	npm install -g @dtoolkit/dops	:7883 (API + MCP), :7884 (Dashboard)

Caution

All server products (dbrain, dwork, dops) require initialization via their init command before the first start. Running start without init will fail.

Typical setup flow

terminal

# 1. Initialize and start the brain
dbrain init && dbrain start

# 2. Initialize and start the project manager
dwork init && dwork start

# 3. Initialize and start observability
dops init && dops start

# 4. Set up context hooks for your AI CLI
dcontext init
dcontext install claude

# 5. Initialize the proxy (optional, for multi-provider access)
dproxy init

For detailed product guides, see the individual product pages: dbrain, dcontext, dwork, dproxy, and dops.