Configuration

Each dtoolkit service stores its configuration in a JSON file under the user’s home directory. Every config value can be overridden with an environment variable, and all services share the same conventions for ports, authentication, and data storage.

Config file locations

Service	Config file	Data directory
dbrain	~/.dbrain/config.json	~/.dbrain/
dwork	~/.dwork/config.json	~/.dwork/
dops	~/.dops/config.json	~/.dops/
dcontext	(per-CLI config)	(none)
dproxy	~/.dproxy/config.json	~/.dproxy/

All config files are created automatically when you run the init command for each service. You can also edit them directly.

terminal

dbrain init    # creates ~/.dbrain/config.json
dwork init     # creates ~/.dwork/config.json
dops init      # creates ~/.dops/config.json
dproxy init    # creates ~/.dproxy/config.json

Service configuration

dbrain

The dbrain config controls the memory server, its storage, networking, and federation connections.

~/.dbrain/config.json

{
  "dataPath": "~/.dbrain",
  "port": 7878,
  "host": "127.0.0.1",
  "token": "your-secret-token",
  "brainType": "personal",
  "connections": [
    {
      "name": "team-brain",
      "url": "http://192.168.1.50:7878",
      "token": "shared-brain-token"
    }
  ]
}

Field	Type	Default	Description
dataPath	string	~/.dbrain	Directory for SQLite database and data files
port	number	7878	REST API and MCP server port
host	string	127.0.0.1	Bind address
token	string	(generated)	Bearer token for API authentication
brainType	"personal" | "shared"	personal	Personal brains are single-user; shared brains accept connections from multiple clients
connections	array	[]	List of remote brains to federate with

Tip

The brainType field determines how the brain behaves. A personal brain is your local memory store. A shared brain runs as a team server that multiple personal brains connect to. When you call recall on a personal brain with connections, it automatically federates the search across all connected shared brains.

Adding connections:

You can add connections via the CLI or by editing the config file directly.

terminal

dbrain connect --name team-brain --url http://192.168.1.50:7878 --token shared-token

dwork

The dwork config controls the project manager server and its storage.

~/.dwork/config.json

{
  "dataPath": "~/.dwork",
  "port": 7881,
  "host": "127.0.0.1",
  "token": "your-secret-token"
}

Field	Type	Default	Description
dataPath	string	~/.dwork	Directory for SQLite index and project data
port	number	7881	REST API and MCP server port
host	string	127.0.0.1	Bind address
token	string	(generated)	Bearer token for API authentication

Tip

dwork uses Markdown files (especially BACKLOG.md) as the source of truth. The SQLite database is only an index for full-text search. You can always regenerate it with dwork sync.

dops

The dops config controls the observability server.

~/.dops/config.json

{
  "dataPath": "~/.dops",
  "port": 7883,
  "host": "127.0.0.1",
  "token": "your-secret-token"
}

Field	Type	Default	Description
dataPath	string	~/.dops	Directory for SQLite database with session data
port	number	7883	REST API and MCP server port
host	string	127.0.0.1	Bind address
token	string	(generated)	Bearer token for API authentication

dcontext

dcontext does not have its own config file. Instead, it writes hooks into each CLI’s native configuration:

CLI	Config modified	Context file
Claude Code	.claude/settings.json	CLAUDE.md
Gemini CLI	settings.json	GEMINI.md
Codex CLI	config.toml	AGENTS.md
OpenCode	npm plugin	AGENTS.md

Install hooks with:

terminal

dcontext install

Or use dbrain’s shorthand:

terminal

dbrain connect claude
dbrain connect gemini

Uninstall hooks to revert all changes:

terminal

dcontext uninstall

Tip

Run dcontext status to see which CLIs have hooks installed and whether dbrain is reachable.

dproxy

The dproxy config controls the universal CLI adapter, including which AI provider to use by default.

~/.dproxy/config.json

{
  "provider": "claude",
  "model": "sonnet",
  "port": 7880,
  "host": "127.0.0.1",
  "token": "your-secret-token"
}

Field	Type	Default	Description
provider	string	claude	Default provider: claude, codex, gemini, or opencode
model	string	(provider default)	Model to use (e.g., sonnet, opus, haiku)
port	number	7880	REST API port (used by dproxy serve)
host	string	127.0.0.1	Bind address
token	string	(generated)	Bearer token for API authentication

Tip

You can override the provider per-request with the --provider flag: dproxy ask --provider gemini "your prompt".

Environment variables

Every config field can be overridden with an environment variable. Environment variables take precedence over values in the config file. The naming convention is {SERVICE}_{FIELD} in uppercase.

dbrain

Variable	Overrides	Example
DBRAIN_DATA_PATH	dataPath	~/.dbrain
DBRAIN_PORT	port	7878
DBRAIN_HOST	host	0.0.0.0
DBRAIN_TOKEN	token	my-secret-token
DBRAIN_BRAIN_TYPE	brainType	shared

dwork

Variable	Overrides	Example
DWORK_DATA_PATH	dataPath	~/.dwork
DWORK_PORT	port	7881
DWORK_HOST	host	0.0.0.0
DWORK_TOKEN	token	my-secret-token

dops

Variable	Overrides	Example
DOPS_DATA_PATH	dataPath	~/.dops
DOPS_PORT	port	7883
DOPS_HOST	host	0.0.0.0
DOPS_TOKEN	token	my-secret-token

dproxy

Variable	Overrides	Example
DPROXY_PROVIDER	provider	gemini
DPROXY_MODEL	model	opus
DPROXY_PORT	port	7880
DPROXY_HOST	host	0.0.0.0
DPROXY_TOKEN	token	my-secret-token

terminal

# Start dbrain on a custom port, bound to all interfaces
DBRAIN_PORT=9000 DBRAIN_HOST=0.0.0.0 dbrain start

Caution

When you set host to 0.0.0.0, the service will be accessible from other machines on your network. Always ensure a strong token is configured before exposing services beyond localhost.

Ports summary

All dtoolkit services use a pair of consecutive ports: one for the API (REST + MCP) and one for the web dashboard.

Service	API port	Dashboard port	Protocol
dbrain	7878	7879	REST + MCP (same port)
dproxy	7880	—	REST + SSE streaming
dwork	7881	7882	REST + MCP (same port)
dops	7883	7884	REST + MCP (same port)

Tip

dbrain, dwork, and dops serve both REST and MCP on the same API port. MCP clients connect to the same URL as REST clients — the server detects the protocol automatically. The dashboard is always served on the next port (API port + 1).

dproxy is the exception: it provides a REST API with SSE streaming but does not have an MCP server or dashboard.

Authentication

All dtoolkit services use Bearer token authentication. A token is generated automatically during init and stored in the config file.

How authentication works

Every HTTP request to a dtoolkit service must include the token in the Authorization header:

terminal

curl -H "Authorization: Bearer your-secret-token" http://localhost:7878/health

MCP clients also authenticate with the same token. When configuring an MCP client (such as Claude Code’s settings.json), pass the token as part of the connection URL or headers.

Setting a token

Tokens are generated automatically when you run init. To set a specific token:

terminal

# Set via the configure command
dbrain configure --token my-new-token

# Or edit the config file directly
# ~/.dbrain/config.json -> "token": "my-new-token"

# Or use an environment variable
DBRAIN_TOKEN=my-new-token dbrain start

Rotating tokens

To rotate a token, update it in the config file or via environment variable, then restart the service. Any connected clients (dcontext hooks, MCP integrations, SDK clients) must be updated with the new token.

terminal

# 1. Generate a new token
NEW_TOKEN=$(openssl rand -hex 32)

# 2. Update the config
dbrain configure --token "$NEW_TOKEN"

# 3. Restart the service
dbrain start

# 4. Update any clients that connect to this service

Caution

For shared brains, token rotation affects all connected clients. Coordinate with your team before rotating a shared brain token. Shared brains also support per-client API keys via dbrain keys for finer-grained access control.

SDK authentication

When using the @dtoolkit/sdk package to connect to services programmatically, pass the token when creating a client:

import { DBrainClient, DWorkClient } from '@dtoolkit/sdk';

const brain = new DBrainClient({
  baseUrl: 'http://localhost:7878',
  token: 'your-secret-token',
});

const work = new DWorkClient({
  baseUrl: 'http://localhost:7881',
  token: 'your-secret-token',
});

Shared brain API keys

Shared brains support multiple API keys for team access control. This lets each team member or service authenticate independently, without sharing a single master token.

terminal

# List existing keys
dbrain keys list

# Create a new key for a team member
dbrain keys create --name "alice"

# Revoke a key
dbrain keys revoke --name "alice"

API keys provide the same access as the master token. The master token (in config.json) always works and cannot be revoked via the keys command.

Next steps

• Getting started — install and initialize your first service
• dbrain — deep dive into the memory server
• dwork — set up your AI-native project manager
• MCP tools reference — full list of MCP tools across all services
• CLI reference — all CLI commands and flags