Riggr

Connector list

Permissions available for every node pair.

Connector list

When you wire an Agent Terminal to another node, the connection's chip lists the available permissions for that pair. This page is the full inventory.

The first option in each list is the default that shows up when the connection is created. Rule of thumb: start restrictive, loosen if needed.

Agent → Code

Connector Grants
code-read Reads the editor content
code-write Replaces the content
code-read-write Reads and writes
code-open-file Opens files in the editor
code-full-access All of the above

Agent → Terminal

Connector Grants
execute Sends commands to the terminal

Agent → Text

Connector Grants
text-read Reads the content
text-write Writes / replaces
text-read-write Reads and writes
summary Logs summaries of actions taken
project-summary Maintains a high-level project summary
commands Logs every Bash command executed
tasks Log of completed tasks
tokens Token consumption report

Agent → JSON

Connector Grants
json-read Reads the JSON
json-write Replaces the JSON
json-read-write Reads and writes

Agent → HTTP Client

Connector Grants
send Fires the request
edit-config Modifies method, URL, headers, body
history Accesses the response history
abort-request Cancels an in-flight request
request-full-access All of the above

Agent → Webview

Connector Grants
open-url Navigates to URLs
screenshot Captures a screenshot
navigate Broad navigation (click, JS injection)
full-access All of the above + form fill, scroll, hover, keys, attributes, arbitrary JS

Agent → Git

Connector Grants
git-commit Creates commits
git-log Reads history
git-full-access All of the above

Agent → File Explorer

Connector Grants
fs-tree Reads the directory tree
fs-search Searches files by name
fs-selection Reads which file is focused
fs-recent-changes Filesystem create / edit / remove events
fs-reveal Highlights a path in the UI
fs-stat Existence, size, mtime, type
fs-full-access All read operations (no create / delete / rename)

Agent → Database

Connector Grants
db-connection-info Driver, host, port, user, database, version (no password)
db-list-databases Databases visible on the connection
db-list-schemas Schemas (Postgres)
db-list-tables Tables, views, collections
db-describe-table Columns, types, defaults, PK, indexes, FKs
db-search-schema Substring search across tables and columns
db-run-query Arbitrary SQL (SQL drivers) or JSON (Mongo)
db-explain EXPLAIN ANALYZE
db-sample-rows SELECT * LIMIT N
db-count-rows COUNT(*)
db-history Last N queries with timings and status
db-full-access All of the above

Agent → NPM Manager

Connector Grants
npm-list-installed Lists packages (deps, devDeps, peer, optional). Read-only — no install or remove.

Agent → Docker

Connector Grants
docker-list Containers, images, volumes, networks
docker-control Start, stop, restart, remove containers
docker-create-container Create new containers
docker-pull-image Pull images
docker-logs-exec Logs and one-shot exec
docker-compose Read, write, manage compose services
docker-full-access All of the above

Agent → Process Manager

Connector Grants
proc-list Lists registered scripts
proc-status Status of a process
proc-logs Last N lines of the buffer
proc-search-logs Grep (substring/regex)
proc-control Start / stop / restart
proc-full-access All of the above

Agent → Environment

Connector Grants
env-read Reads variables (won't expose secrets)
env-add Adds a variable
env-edit Edits an existing variable
env-full-access All of the above

Agent → Task

Connector Grants
task-set-schedule Sets cron/seconds and prompt
task-control Start / stop / status
task-delegate-and-fire Schedules a future fire; agent exits; when time arrives, prompt returns as a new message

Agent → Agent (delegation)

Connector Grants
agent-delegate Sends prompt, waits for reply
agent-send Fire-and-forget (send and move on)
agent-read-output Reads the other agent's output
agent-interrupt Remote Ctrl+C
agent-full-access All of the above

Automatic typed-port integrations

When two nodes have compatible ports, the chip shows a ports:<output>-><input> integration. The agent isn't involved — direct dataflow.

Port connector From → To What it does
ports:output->appendText Terminal → Text Each terminal line becomes a paragraph in the text
ports:consoleErrors->appendText Webview → Text Browser console errors become text
ports:currentUrl->url Webview → HTTP Client Current URL becomes the request URL
ports:consoleErrors->restartAll Webview → Process Manager Console error restarts every process
ports:containerLogs->appendText Docker → Text Container logs land in the text
ports:variables->applyEnv Env → Process Manager Vars become each process's env
ports:variables->headers Env → HTTP Client Vars become request headers
ports:lastResult->setJson Database → JSON Last result shows up as queryable JSON
ports:lastResultMarkdown->setContent Database → Text Result becomes a markdown table
ports:queryHistory->appendText Database → Text Query history flows into text
ports:json->body JSON → HTTP Client JSON becomes the request body
ports:lastResponse->setJson HTTP Client → JSON Last response as an object
ports:lastResponseText->content HTTP Client → Code Last response as text
ports:selectedFile->file File Explorer → Code Selected file opens in the editor
ports:selectedFile->loadFile File Explorer → Text Selected file loads into the text
ports:recentChanges->appendText File Explorer → Text Filesystem changes become text
ports:savedAt->refresh Code → Git Saved a file? Git refreshes
ports:crashes->prompt Process Manager → Agent Terminal Crash fires an automatic prompt
ports:logs->appendText Process Manager → Text Log stream
ports:crashes->appendText Process Manager → Text Crash log
ports:runningCount->setContent Process Manager → Text "N/M running" counter
ports:prompt->prompt Task → Agent Terminal Scheduler fire becomes a prompt in the agent

When a combination doesn't show up

If the chip shows no options, that node pair doesn't have a native integration yet. The way around it is to give the agent a route it already has (usually the Terminal) and describe in the prompt what you want done via command.