Command Reference

Detailed reference for every slash command in Glean Code. Looking for a quick overview? See the README's Commands at a glance.

Commands marked Mock mode below read from the built-in fake corpus when no credentials are configured — seventy interlinked documents spread evenly across five faux datasources (gdrive, confluence, jira, github, slack — 14 each). Full inventory and the people roster: Mock Corpus.

/help

Show help for a specific command, or list every available command grouped by category.

/help [command]

Parameter	Description
`command`	Optional. Command name without the leading `/`. Omit to list all commands.

/help
/help search
/help agents.run

Output — Without an argument: a grouped list of all commands with one-line summaries. With a command name: usage signature, parameter table, examples, and the underlying REST endpoint.

Endpoint — (local)

/status

Show the current connection state, active mode, configured instance, and any impersonation in effect.

/status

/status

Output — A summary box showing: mode (live / mock / auto), instance host, token presence, act-as email if set, and the active chat thread id.

Endpoint — (local)

/doctor

Run a full health check: validates config, tests DNS resolution, TCP connectivity, and performs a live auth probe against the search endpoint.

/doctor

/doctor

Output — A line per check (config, url shape, dns, tcp, auth probe) with a pass/fail status and detail message. Useful for diagnosing connectivity or token problems.

Endpoint — (local checks + POST /rest/api/v1/search probe)

/login

Sign in to Glean with browser OAuth, or store an API token. OAuth uses Authorization Code + PKCE and Dynamic Client Registration when no static client ID is configured.

/login <hostname-or-instance-id> [--client-id <id>] [--port <n>] [--no-browser] [--act-as <email>]

Parameter	Description
`hostname-or-instance-id`	A backend hostname such as `acme-be.glean.com`, or an instance ID such as `acme`. IDs map to `<id>-be.glean.com`.
`--instance`	Alternative named form of the hostname or instance ID.
`--client-id`	Optional static OAuth client ID. DCR is used when omitted.
`--port`	Optional fixed localhost callback port for redirect-URI allowlisting.
`--no-browser`	Print the OAuth authorize URL instead of opening a browser.
`--token`	Legacy API-token login. Supplying it skips OAuth.
`--act-as`	Optional. Email address to impersonate via `X-Glean-ActAs`.

/login acme
/login acme-be.glean.com
/login --instance acme-be.glean.com --no-browser
/login --instance acme-be.glean.com --token glean_tok_xxx

Output — OAuth opens the browser and waits for the localhost callback, then stores OAuth tokens in ~/.gleancode/auth.json. Token login stores the API token in ~/.gleancode/config.json.

Endpoint — (local, OAuth Authorization Code + PKCE or Authorization header)

/logout

Clear stored credentials and revert to mock mode.

/logout

/logout

Output — Confirms credentials removed.

Mock mode — This is what /logout reverts to: with no token, searches answer from the mock corpus. It clears both API-token and OAuth credentials.

Endpoint — (local)

/open

Open the configured Glean instance in the default web browser. Strips the -be suffix from the instance host, so acme-be.glean.com opens https://acme.glean.com.

/open [path] [--path <path>] [--print]

Parameter	Description
`path`	Optional path to append, positionally. e.g. `search` or `/docs/123`. A leading `/` is added if missing.
`--path`	Same as the positional path. Either form works.
`--print`	Print the URL instead of launching a browser. Useful over SSH. `--dry-run` does the same.

/open
/open search
/open /docs/123
/open --print

Output — The URL being opened, then the browser launch. With --print: just the URL. Errors if no instance is configured.

Endpoint — (local)

/config

View or update individual configuration keys. Changes are persisted to ~/.gleancode/config.json.

/config [get <key> | set <key> <value> | list]

Subcommand	Description
`list`	Print the full config as key/value pairs.
`get <key>`	Print the value of a single key.
`set <key> <value>`	Update a key. See the Config keys section for valid keys and values.

/config list
/config get mode
/config set mode live
/config set default_page_size 20
/config set indexing_token glean_idx_xxx
/config set mock_corpus_path ~/demo-corpus.json

Output — For list and get: the value(s). For set: a confirmation message.

Mock mode — mock_corpus_path points offline mode at your own JSON corpus instead of the built-in Acme one; see Bring your own corpus. Unset it with /config set mock_corpus_path "".

Endpoint — (local, ~/.gleancode/config.json)

/mode

Quickly switch the API mode without editing config.

/mode <live|mock|auto>

Value	Description
`live`	Force all API calls to the real Glean backend.
`mock`	Force all calls to return local fake data (no network).
`auto`	Use live if credentials are present, otherwise fall back to mock.

/mode auto
/mode mock
/mode live

Output — Confirms the new mode.

Endpoint — (local)

/mcp

Inspect, configure, and run the bundled MCP server without leaving the REPL.

/mcp <status|config|start|stop> [client] [--url] [--stdio] [--name <key>] [--port <n>] [--host <addr>] [--transport <t>] [--mock]

Subcommand	Description
`status`	Installed `mcp` version and whether it can run the server, plus the running server's pid, URL, uptime, and mode. The default when no subcommand is given.
`config`	Print the JSON block to paste into a client. Optional client name: `claude-code`, `claude-desktop`, `cursor`.
`start`	Run the server detached over HTTP and record where it landed.
`stop`	Terminate the server `start` launched.

Flag	Description
`--url`	`config`: emit the URL form pointing at the running server.
`--stdio`	`config`: force the command form even while a server is running.
`--name`	`config`: key under `mcpServers`. Default `glean`.
`--port`	`start`: bind port. Default `8787`.
`--host`	`start`: bind address. Default `127.0.0.1`.
`--transport`	`start`: `streamable-http` (default) or `sse`.
`--mock`	`start`: serve the built-in corpus. Implied when the REPL is in mock mode.

/mcp status
/mcp config claude-code
/mcp config --name glean-cli
/mcp start --port 9000
/mcp config --url
/mcp stop

Output — status prints a key/value table; config prints JSON; start and stop confirm with the pid and URL.

start cannot use stdio, by design. stdio is the transport an MCP client spawns for itself over a pipe pair — started from the REPL it would have no client on the other end. /mcp start therefore runs streamable-http (or sse), which a client attaches to by URL. For the ordinary setup, use /mcp config and let the client spawn the server. See docs/MCP.md.

Name collisions. The emitted block keys the server as glean under mcpServers. Glean's own hosted MCP server would naturally be registered under that name too, and pasting over it silently swaps the toolset with no error. Use --name glean-cli to run both side by side.

Endpoint — (local — spawns glean_mcp.py)

/flow

Map what you have investigated: captured chats, the documents they cited, and the connections between them. Full guide: docs/FLOW_MAPPER.md.

/flow <status|enrich|link|show|timeline|purge> [--docs <n>] [--links <n>] [--limit <n>] [--min-score <f>] [--output <file>] [--print] [--all] [--older-than <days>]

Subcommand	Description
`status`	Capture setting, database path and size, and what has been recorded. The default when no subcommand is given.
`enrich`	Fetch document text for captured citations, via `/getdocuments` then `/summarize`. Required before linking.
`link`	Find document-to-document and cross-session links.
`show`	Draw the captured investigations as a vertical rail, with connections branching off it.
`timeline`	Write a self-contained HTML timeline and open it.
`purge`	Delete captured data, after confirming.

Flag	Description
`--docs`	`show`: documents listed per session before the rest are counted. Default `6`.
`--links`	`show`: connections drawn per session before the rest are counted. Default `3`.
`--limit`	`enrich`: documents to fetch in one run. Default `50`.
`--min-score`	`link`: phrase-link threshold, 0–1. Default `0.45`. Higher is stricter.
`--output`	`timeline`: where to write the HTML. Default a temp file.
`--print`	`timeline`: write without opening a browser.
`--all`	`purge`: every instance and mode, not just the current one.
`--older-than`	`purge`: only sessions older than this many days.

/flow status
/flow enrich
/flow link --min-score 0.6
/flow show
/flow show --docs 3
/flow timeline --output ~/flow.html
/flow purge --older-than 30

Output — show draws each session as a node on a vertical rail, with its questions, sources, and any connection branching off in yellow:

── flow: acme-be.glean.com · mock ────────────────────────────────────────────

  ●─ 1  what happened in the checkout incident?
  │     Tue 18 Aug 2026, 22:05  ·  4 turns  ·  6 sources
  │     ↳ who owned the fix?
  │
  │     ▪ confluence Postmortem: Checkout Latency Incident (INC-1183)
  │     ▪ slack      War room thread: checkout 5xx spike
  │     ▪ jira       INC-1183 — Elevated 5xx on checkout API
  │     … 3 more documents
  │
  ├──◆ linked-document  0.60  →  2 ───────────────────────────────────────────
  │    Postmortem: Checkout Latency Incident (INC-1183)
  │      ↓  shares: incident, checkout
  │    Customer QBR — Northwind Retail
  │
  ●─ 2  what are the risks going into the Northwind renewal?
  │     Tue 18 Aug 2026, 22:05  ·  2 turns  ·  3 sources
  │
  │     ▪ gdrive     Customer QBR — Northwind Retail
  │     ▪ jira       SUP-882 — Northwind: search results missing Confluence…
  │     ▪ slack      Northwind attachment issue — need connector eyes
  │
──────────────────────────────────────────────────────────────────────────────

Documents a thread kept returning to lead the list; the rest hold the order they were cited in. Connections are ordered by how much they tell you — a linked-document link found something, while a shared-citation link between two runs of the same question is trivially certain and says little — so the discovery is never buried under a wall of 1.00 scores.

Each document is tagged with its datasource in a consistent colour, so a source is recognisable before you read its name. A connection names the session it reaches (→ 2), stacks the two documents that bridge around a ↓, and prints the shared evidence — shares: incident, checkout — so every link can be read rather than taken on trust. Colour is decoration only: piped, redirected, or under NO_COLOR, the glyphs still carry the structure.

status prints a table; timeline reports the file written; purge confirms before deleting.

Capture is opt-in for live data. The flow_capture config key defaults to mock, so real tenant content is never recorded until you set it to on. A local database has no permission model — see the retention section before enabling it against a tenant.

Mock mode — the built-in corpus is what this feature was tuned against: seven identifier clusters plus a QBR that references an incident in prose with no ticket number, which is the link worth finding.

Endpoint — (local — ~/.gleancode/flow.db; enrich calls /getdocuments or /summarize)

/ask

Translate a natural-language request into a sequence of Glean Code slash commands using Glean Assistant as the planner. Read docs/NATURAL_LANGUAGE.md for the full design.

/ask "<natural-language-request>"
?<natural-language-request>

The ? prefix is shorthand. Both forms are identical — ?login into acme-be.glean.com is the same as /ask "login into acme-be.glean.com".

Parameter	Description
`<request>`	Plain English describing what you want to do. Multi-step requests work — "do X, then Y, and finish with Z".

/ask "login into acme-be.glean.com with the stored token"
?login into acme-be.glean.com, then search for "Q2 plan"
?show me datasource health and start a chat
?rotate my indexing token and show its status

How it works — The CLI sends your request to /chat with a system prompt asking Glean to emit a JSON array of {cmd, args} steps. The reply is parsed locally, validated against the registered command set, rendered as a numbered plan, and (if any step is destructive) gated behind a single Run all? [y/N] confirm. Reads run automatically. Mock mode emits a canned plan without calling Glean.

Output — A numbered plan with [confirm] or [unknown — will be skipped] markers, then the output of each executed step.

Endpoint — POST /chat against your Glean instance. Full architecture in docs/NATURAL_LANGUAGE.md.

/history

Show commands entered during the current session.

/history [--limit <n>]

Parameter	Description
`--limit`	Maximum number of entries to show. Default `20`.

/history
/history --limit 5

Output — A numbered list of recent commands, newest last.

Endpoint — (local)

/clear

Clear the terminal screen.

/clear

/clear

Endpoint — (local)

/exit

Quit Glean Code.

/exit

/exit

Endpoint — (local)

/chat

Send a message to the Glean Assistant. Continues the current thread by default; use --new to start a fresh conversation.

/chat <message> [--new] [--chat-id <id>] [--agent <name>] [--stream]

Parameter	Description
`message`	The user message. Quote it if it contains spaces.
`--new`	Start a new chat thread, discarding the current thread id.
`--chat-id`	Continue a specific chat thread by id.
`--agent`	Route the turn through a named agent configuration.
`--stream`	Request a streaming response.

/chat "what did engineering ship last week?"
/chat "summarise the Q2 plan" --new
/chat "draft an email to Alice" --agent sales

Output — The assistant's response in a styled box, with cited source documents listed below. The active chat thread id is saved for subsequent /chat calls.

Mock mode — Citations are ranked against the mock corpus, so every cited document is one /docs.get and /summarize can fetch.

Endpoint — POST /rest/api/v1/chat

/search

Search the Glean index and display ranked results with snippets.

/search <query> [--page-size <n>] [--datasource <name>]

Parameter	Description
`query`	Free-text search query.
`--page-size`	Number of results to return. Defaults to `default_page_size` in config.
`--datasource`	Restrict results to one datasource, e.g. `gdrive`, `slack`, `jira`, `confluence`.

/search "quarterly planning"
/search "oncall runbook" --datasource confluence --page-size 5

Output — Numbered result list, each showing title, datasource, URL, a byline (author, document type, container, freshness) when the API returns metadata, and a matching text snippet.

Mock mode — Results are ranked against the mock corpus: query terms score against title, tags, body, and container, --datasource genuinely filters (gdrive, confluence, jira, github, slack), and snippets come from the sentence that best matches the query. A query that matches nothing is padded with the freshest documents rather than returning empty. Page size is capped at 10.

Endpoint — POST /rest/api/v1/search

/datasources.list

List all datasources visible to the current token, derived from a faceted search call.

/datasources.list [--with-counts] [--with-status] [--sample <n>]

Parameter	Description
`--with-counts`	Show document counts per datasource.
`--with-status`	Fetch full indexing status per datasource (requires `indexing_token`).
`--sample`	Sample size for the underlying search call. Default `100`.

/datasources.list
/datasources.list --with-counts
/datasources.list --with-status
/datasources.list --with-counts --sample 200

Output — A list of datasource names. With --with-counts: document counts alongside each name. With --with-status: uploaded/indexed counts and coverage % from the Indexing API.

Mock mode — Returns the five faux datasources with plausible index counts: gdrive 1840, confluence 920, slack 611, jira 430, github 268. See Mock Corpus.

Endpoint — POST /rest/api/v1/search (facets) + POST /api/index/v1/debug/{ds}/status

/autocomplete

Get query suggestions for a partial search string.

/autocomplete <partial>

Parameter	Description
`partial`	The in-progress query string.

/autocomplete "quart"
/autocomplete "onboard"

Output — A list of suggested completions ranked by relevance.

Mock mode — Completions are built from the tags of documents matching what you typed, e.g. /autocomplete "quart" → quart planning, quart tracker, quart capacity.

Endpoint — POST /rest/api/v1/autocomplete

/recommendations

Get document recommendations for a user based on their activity and context.

/recommendations [--user <email>]

Parameter	Description
`--user`	Target user email. Defaults to the authenticated user.

/recommendations
/recommendations --user alice@acme.com

Output — A list of recommended document titles and URLs.

Mock mode — The five freshest mock corpus documents, in full search-result shape.

Endpoint — POST /rest/api/v1/recommendations

/feedback

Send explicit feedback on a search result or chat turn using its tracking token.

/feedback <tracking-token> <rating> [--comment <text>]

Parameter	Description
`tracking-token`	The `trackingToken` field returned in a search result or chat response.
`rating`	`THUMBS_UP` or `THUMBS_DOWN`.
`--comment`	Optional free-text comment.

/feedback tok_1 THUMBS_UP
/feedback tok_1 THUMBS_DOWN --comment "wrong datasource"

Output — Confirmation that feedback was recorded.

Endpoint — POST /rest/api/v1/feedback

/datasources.status

Show full indexing status for a single datasource: visibility, document upload and index counts, coverage percentage, and the last five processing events.

Requires indexing_token — set it with /config set indexing_token <token>.

/datasources.status <datasource>

Parameter	Description
`datasource`	Datasource name, e.g. `slack`, `gdrive`, `confluence`.

/datasources.status slack
/datasources.status gdrive

Output — Datasource visibility, uploaded document count, indexed document count, coverage %, and a table of the last 5 processing history events with timestamps.

Endpoint — POST /api/index/v1/debug/{datasource}/status

/indexing.rotate-token

Rotate the indexing API token secret and print the new raw secret. Store it immediately — the old secret is invalidated.

/indexing.rotate-token

/indexing.rotate-token

Output — The new raw secret and a reminder to run /config set indexing_token <new-secret>.

Endpoint — POST /api/index/v1/rotatetoken

/datasources.config

Get the live configuration for a custom datasource — object definitions, ACL settings, trusted domains, icon URL.

/datasources.config <datasource>

/datasources.config gdrive
/datasources.config custom1

Output — Datasource config object (object types, isUserReferencedByEmail, trustedDomains, datasourceCategory, etc.).

Endpoint — POST /api/index/v1/getdatasourceconfig

/documents.status

Get upload + indexing status for a single document.

/documents.status --datasource <ds> --object-type <type> --id <doc-id>

/documents.status --datasource gdrive --object-type Article --id doc-1

Output — uploadStatus, lastUploadedAt, lastIndexedAt, any indexing errors.

Endpoint — POST /api/index/v1/getdocumentstatus

/documents.count

Count uploaded documents in a custom datasource.

/documents.count --datasource <ds>

/documents.count --datasource custom1

Output — Document count.

Endpoint — POST /api/index/v1/getdocumentcount

/users.count

Count users uploaded for a custom datasource.

/users.count --datasource <ds>

/users.count --datasource custom1

Output — User count.

Endpoint — POST /api/index/v1/getusercount

/documents.access

Check whether a specific user has access to a specific document — useful for debugging "why can't user X see doc Y?"

/documents.access --datasource <ds> --object-type <type> --id <doc> --user <email>

/documents.access --datasource gdrive --object-type Article --id doc-1 --user alice@example.com

Output — YES / NO access decision.

Endpoint — POST /api/index/v1/checkdocumentaccess

/debug.document

Get debug info (status + uploaded permissions) for a single document.

/debug.document <datasource> <doc-id> [--object-type <type>]

/debug.document gdrive doc-1 --object-type Article

Output — Document upload status, last-uploaded/indexed timestamps, ACL permissions.

Endpoint — POST /api/index/v1/debug/{datasource}/document

/debug.documents

Bulk debug for multiple documents in a datasource.

/debug.documents <datasource> --from-file <items.json>

The JSON file should be an array of {objectType, docId} entries.

/debug.documents gdrive --from-file ./batch.json

Output — Per-document debug results.

Endpoint — POST /api/index/v1/debug/{datasource}/documents

/debug.user

Get debug info for a user in a datasource — upload status + uploaded group memberships.

/debug.user <datasource> <email>

/debug.user gdrive alice@example.com

Output — User upload status and groups uploaded via the permissions API.

Endpoint — POST /api/index/v1/debug/{datasource}/user

/index.document

Index a single document. Supports two input modes — pass a fully-formed JSON body, or point at a local file and let the CLI synthesize the body for you.

/index.document (--from-file <doc.json> | --path <file>) \
                [--datasource <ds>] [--object-type <type>] \
                [--public | --acl-from-file <perms.json>] \
                [--id-prefix <s>] [--view-url-prefix <url>] \
                [--version <n>] [--dry-run]

Flag	Description
`--from-file`	JSON file containing a full `DocumentDefinition` body
`--path`	Local file path. Synthesizes a `DocumentDefinition` from the file. Errors if a directory is passed — use `/index.bulk-documents --path` for folders
`--datasource`	(with `--path`) Datasource name
`--object-type`	(with `--path`) Object type (e.g. `Article`)
`--public`	(with `--path`) Make the doc world-readable. Mutually exclusive with `--acl-from-file`
`--acl-from-file`	(with `--path`) JSON file containing a `DocumentPermissionsDefinition` applied to the doc
`--id-prefix`	(with `--path`) String prepended to the path-derived id slug
`--view-url-prefix`	(with `--path`) Base URL prepended to the relative path. Defaults to `file://<absolute>`
`--version`	Optional version int for optimistic concurrency
`--dry-run`	Print the assembled request body and exit without calling the API

Supported file types when using --path: .txt, .md, .markdown, .html, .htm, .json. Binary formats (PDF, .docx) are out of scope.

/index.document --from-file ./doc.json
/index.document --path ./README.md --datasource custom1 --object-type Article --public
/index.document --path ./team/onboarding.md --datasource custom1 --object-type Article \
                --acl-from-file ./perms.json --dry-run

Output — Acceptance status from the API. With --dry-run, the assembled JSON body.

Endpoint — POST /api/index/v1/indexdocument