Appearance
ts ai
Administer the AI pipe from the command line: the model endpoints requests go out through, the tools and MCP connectors a model is allowed to reach, the stored agent and skill definitions, and the observability behind all of it.
Usage
text
ts ai
ts ai <command> [<args>]Description
A bare ts ai lists the namespace instead of erroring on a missing subcommand, so it doubles as the reminder of what lives under here.
Every noun comes as a pair. The plural lists — ts ai endpoints prints the table. The singular acts on one row — ts ai endpoint <name> prints one endpoint, and creates, changes, or deletes it.
Writes share one grammar: ts ai <noun> set <name> <property> <value>, with kebab-case property names. The name is whatever the plural listing shows in NAME; the row's id works too. A name that two rows share is refused rather than guessed at, because acting on the wrong endpoint or connector silently repoints live traffic.
Secrets are write-only throughout. set rotates one, and nothing anywhere reads one back — listings show only that a secret is configured, and sometimes a last-four cue.
Everything under ts ai requires a Server Administrator account, and every write is recorded in the audit log. See ts audit for the record.
In This Section
| Command | Description |
|---|---|
| ts ai endpoints / ts ai endpoint | Model provider connections. The only noun here you can create from ts. |
| ts ai tools / ts ai tool | The actions a model may invoke, globally and per workspace. |
| ts ai connectors / ts ai connector | MCP servers the pipe can reach, and the tokens it reaches them with. |
| ts ai agents / ts ai agent | Stored agent definitions. Enablement only. |
| ts ai skills / ts ai skill | Stored skill definitions. Enablement only. |
| ts ai transcripts / ts ai transcript | Per-call metadata, and the audited read of one call's content. |
| ts ai usage | Calls, errors, and token totals by workspace, provider, and model. |
Related Commands
- ts audit — Where every write in this namespace lands, including who did it.
- ts workspace — The workspaces tools and connectors are enabled for.
- ts usage — Session and user usage, which is a different question from token spend.