The agents view

When several agents work for you at once, you need one place to see them. Syntic Code has two:

The agents dashboard (syntic agents) is a full-screen list of every background session on your computer, across all projects. Use it from any terminal to check on long jobs, read their logs, step into one, or clean up.
The in-session agents view (← on an empty prompt, or /tasks) lists everything working for the session you are in: the main conversation, its subagents and the background work it started.

Quick start

syntic --bg "upgrade the test suite to the new fixtures API and get it green"
syntic --bg "write docs for every public function in src/export"
syntic agents

 STATE     ID        PROMPT                                     FOLDER          STARTED   COST    TOKENS
› running  08e14168  upgrade the test suite to the new fixtur…  ~/code/shop     4m ago    $0.38   212.4k
  finished 5c1d0a2e  write docs for every public function in …  ~/code/shop     3m ago    $0.21   96.0k

Select a row and press Enter to continue that session yourself, or l to read its log.

The agents dashboard

syntic agents opens the dashboard. It re-reads the list every two seconds, so new runs and state changes appear on their own.

What each row shows

Column	Meaning
`STATE`	`running`, `finished`, `failed`, `stopped`, `interrupted` or `needs input`. A run whose process vanished without finishing shows as failed.
`ID`	The run’s short id. Any unique prefix works in the shell commands.
`PROMPT`	The start of the task you gave it.
`FOLDER`	Where it was started.
`STARTED`	How long ago it began.
`COST`, `TOKENS`	What it has used so far, read from its saved conversation.

The state meanings are on Background sessions.

Keys

Key	Does
`↑` `↓` (or `k` `j`), `Home`, `End`	Move the selection.
`Enter` or `a`	Attach: leave the dashboard and open that session in the interactive screen, so you can continue it.
`l` or `→`	Open the run’s log.
`s`	Stop the selected run. You confirm with `y`; it is asked to stop, then forced after a few seconds.
`p`	Prune: remove every finished, failed and stopped run and its log, after a `y` to confirm. Running sessions are never removed.
`r`	Refresh now.
`q`, `Esc` or `Ctrl+C`	Close the dashboard.

In the log view:

Key	Does
`↑` `↓` (or `k` `j`), `PgUp`, `PgDn`	Scroll. The view follows new output until you scroll up.
`End` or `G`	Jump back to the newest output and follow it.
`a` or `Enter`	Attach to this run.
`s`	Stop this run.
`Esc`, `q`, `←` or `l`	Back to the list.

Options

Option	Does
`--cwd <path>`	Only runs started in that folder or below it.
`--json`	Print the runs as JSON and exit, for scripts and status bars.
`--list`	List your custom agent definitions instead of the dashboard.

When syntic agents is not run in a terminal (piped, or in a script), it prints the agent definitions, the same as --list.

Read the list from a script

syntic agents --json | jq -r '.[] | select(.state == "needs-input") | .id'

Each entry has these fields:

Field	Meaning
`id`	The background run id.
`session_id`	The conversation behind it, for `syntic --resume`.
`state`	`running`, `finished`, `failed`, `stopped`, `interrupted` or `needs-input`.
`prompt`, `cwd`	The task and the folder.
`started_at`, `ended_at`, `exit_code`	Timing and how it ended.
`pid`	The process id, while running.
`log`	Path to its log file.
`cost_usd`, `input_tokens`, `output_tokens`	Usage so far. Input tokens count fresh input only, not cache reads or writes.

The same from the shell

Every dashboard action has a command, useful over SSH or in scripts:

Command	Does
`syntic sessions` (or `syntic ps`)	List runs.
`syntic logs <id> [-f]`	Print the log; `-f` follows it.
`syntic attach <id>`	Continue the run in this terminal.
`syntic stop <id>`	Stop it.
`syntic respawn <id>`	Restart an interrupted, failed, stopped or waiting run in the background.
`syntic rm <id>`	Remove a finished run.
`syntic sessions --prune [--older-than <days>]`	Remove old runs.

The in-session agents view

Inside a session, press ← with the prompt empty, or run /tasks, to open the agents view. It lists the main session first, then each subagent and background agent this session started, with the running ones on top:

❯ ◉ main session      fix the flaky parser test            working · 1m 04s · 18.2k tokens
  ● Explore           find every caller of parse_header    running · 32s · 1 queued
  ✔ general-purpose   write the migration tests            done · 2m 10s · 9.4k tokens

Each row shows the agent type, its task, its state, how long it has run and what it has used. A count of queued messages appears when other agents have sent it messages it has not read yet. Agents from earlier runs of a resumed session are listed too.

Key	Does
`↑` `↓`, `PgUp` `PgDn`	Select.
`Enter`	Open that agent’s conversation (on the main session row, go back to the prompt).
`/`	Filter the list by typing; `Enter` keeps the filter, `Esc` clears it.
`s`	Cycle which states are shown.
`x` `x`	Stop the selected running agent (press `x` twice within three seconds).
`Esc`	From a conversation, back to the list; from the list, back to the prompt.

Ctrl+X then Ctrl+K stops every background agent at once, from anywhere in the session.

Open it automatically

{ "showAgentsViewByDefault": true }

With this setting, the view opens by itself when background agents start.

Other places agents show up

A live card under each Agent(...) row in the transcript shows the subagent’s current activity, time and tokens, then folds to one line when it finishes.
The agent panel under the footer lists running agents at a glance.
/agents manages custom agent definitions (create, edit, delete) and has a Running tab with the same live list.

Where the state lives

Background runs are recorded in a background folder inside Syntic Code’s data folder (on macOS, ~/Library/Application Support/ai.syntic.syntic/background/), with one log per run. Each run’s conversation is saved like any other session, so syntic --resume <session_id> also works. See The .syntic folder.

Troubleshooting

Problem	Cause and fix
`syntic agents` prints agent definitions instead of the dashboard	It is not attached to a terminal (piped or in a script). Run it in an interactive terminal, or use `--json`.
The dashboard says there are no background sessions	Only runs started with `syntic --bg`, `/background` or `/branch --bg` appear. Subagents of an open session are in that session’s agents view (`←`).
A run shows `needs input`	It asked for a permission that could not be granted unattended. `syntic attach <id>` to answer, or start such runs with the permissions they need (`--allowed-tools`, `--permission-mode acceptEdits`).
A run shows `interrupted`	Its process ended without finishing, for example after a restart. `syntic respawn <id>` continues it; `syntic respawn --interrupted` continues all of them.
Disk filling with old runs and worktrees	Press `p` in the dashboard, or `syntic sessions --prune --older-than 7`.

Subagents Background sessions