List the query history

tobira history lists the queries that tobira query, the terminal UI’s editor, the AI assistant’s tools and MCP clients ran, newest first, and tobira history clear removes entries. What is stored, where, for how long and how credentials are masked is on the Query history page.

tobira history
tobira history --conn local-pg --search "from orders" --limit 20
tobira history --mcp --format json
tobira history clear --search "hunter2" --yes

tobira history

tobira history [--conn <name>] [--source cli|tui|ai|mcp] [--mcp] [--search <text>] [--limit <n>] [--format table|json|ndjson|csv]

Each entry has at, conn, source (cli, tui, ai or mcp), ok, rows, ms and query. json and ndjson also carry client (null unless the call came from an MCP client), and table and csv add a client column for an MCP listing.

Flag	Default	Meaning
`--conn <name>`	all	Only entries of this connection.
`--source <src>`	all	Only entries from one source: `cli`, `tui`, `ai` or `mcp`. Any other value is a usage error.
`--mcp`		Only calls made by MCP clients, with the client’s name. The same as `--source mcp`.
`--search <text>`		Only entries whose query contains the text, ignoring case (non-ASCII text too).
`--limit <n>`	`100`	Maximum entries; `0` lists every retained entry (up to `history.max_entries`, 10,000 by default). A negative value is a usage error.
`--format <fmt>`	`table` on a terminal, else `json`	Same formats as `tobira query`.

With no history yet it prints nothing ([] for json) and exits 0.

clear

tobira history clear [--conn <name>] [--source cli|tui|ai|mcp] [--mcp] [--search <text>] [--yes]

Removes every entry, or only the entries that match the filters, which mean the same as for the listing: --conn, --source, --mcp and --search. On a terminal it shows how many entries match and asks you to type yes; without a terminal --yes is required, and without it nothing is removed (exit 1). It prints removed <n> entries, and removed 0 entries when nothing matches or there is no history file yet, which it never creates. The file is compacted afterwards, so a removed query is gone from the disk as well.