CLI

fb <command> [args] [flags]

fb <command> --help is authoritative, because it is generated from the same operation definitions the commands run on. This page is the map.

Every read command takes exactly one reference. There is no multi-argument form: fb page nasa zuck is a usage error, and a shell loop is the answer.

A reference is a handle, a numeric id, a permalink, a share link, a story key, or any Facebook URL fb can classify. fb id <thing> says which of those it is without making a request.

Reads

Command	What it emits
page <handle|id>	A Page or profile whole: identity, counts, contact, tabs, the timeline's first post
profile <handle|id>	The same command under the other name
feed <handle|id>	The timeline. One post signed out, paged with a session
post <url|id>	A post, with its message entities, counts, attachments and the comments the permalink ships
comments <url|id>	Just the comments from that read
reactions <url|id>	Just the reaction breakdown, all seven types
photos <handle|id>	The photo tab, or an album with --album
photo <fbid|url>	One photo: image, alt, owner, album, caption, counts, comments, next
videos <handle|id>	The video tab, grid and shows stitched together
video <id|url>	One video: media URLs, dimensions, captions, music
reel <id|url>	The same command under the other name
events <handle|id>	The events tab, past and upcoming
event <id|url>	One event: host, time, place, description, RSVP counts
group <id|slug>	A group's public shell
group feed <id|slug>	A group's discussion
discover	The Pages directory, as a seed source
search <query>	Search. Needs a session, exit 4 without one

Flags of their own

page      --about        also fetch the About page
          --no-posts     skip the timeline parse
          --tab          read one tab instead of the main page
post      --author       the profile a bare post id or a pfbid belongs to
comments  --author       the same
photos    --album        walk the album this photo is in, not the photo tab
          --download     write every image and its sidecar into this directory
photo     --download     the same, for one photo
video     --transcript   also fetch the watch route, which carries the transcript
          --download     write the video and its sidecar, HD if there is one
videos    --playlists    print the shows rather than the videos
event     --suggested    also print the events Facebook suggests alongside
group     --resolve-opaque   spend a request each on the pfbid authors in a feed

Claims

Command	What it emits
edges <ref>	The claims one read already makes. No request of its own beyond that read
graph <ref>	edges plus a walk. --depth hops, --budget requests
rdf <ref>	The claims as RDF. --format nt|turtle|jsonld, --depth, --budget, --no-provenance

Stores

Command	What it emits
crawl <seed>...	Walks from the seeds into a store, then a manifest line
db stats	What is in the store, counted as nodes, claims and reads
query <sql>	Plain SQLite over the store, opened read-only
export	The whole store as RDF. --format, --no-provenance

crawl flags: --depth, --budget, --store, --seed-directory, --letter, --resolve-opaque.

--store is the store these four commands read and write. --db is the global tee flag, which writes records into a store as a side effect of any read. They are separate on purpose.

Captures and surfaces

Command	What it emits
archive <ref>	One read written down whole into --dir. Never from the cache
serve	Every read operation over HTTP as NDJSON. --addr, --allow-writes
mcp	The same operations as MCP tools over stdio

Questions about fb

None of these five make a request.

Command	What it emits
id <anything>	What a string is, which command handles it, and what is still uncertain. --resolve spends one request
explain <url>	id plus the routing: surface, GraphQL operation, tier, depth
surfaces	The eight surfaces and what each answers
tiers	The two tiers and what each needs
routes	Command to surface to operation, at each tier. --questions keys it by what you want to know instead
fields [kind]	Every field on a record kind, and how often the captured pages filled it

Session

Command	What it does
auth import --c-user <id> --xs <token>	Store the two cookies
auth status	Whether a session is stored, and whose. No request
auth logout	Forget it
whoami	Prove the stored session against Facebook. One request

Housekeeping

Command	What it does
cache	What is in the page cache
cache clear	Delete it
config show	The resolved configuration
config path	The data directory
completion <shell>	Shell completion

Global flags

Every command takes all of these.

-o --output      auto|table|markdown|list|json|jsonl|csv|tsv|url|raw
   --fields      which columns to show
   --template    a Go template applied per record
   --no-header   omit the header row
   --color       auto|always|never
-n --limit       stop after N records (0 = no limit)
   --tier        cap the tier (0|1) or pin one surface
   --rate        minimum delay between requests
   --retries     retry attempts on rate limit or 5xx
   --timeout     per-request timeout
   --proxy       an HTTP proxy for every request
   --no-cache    bypass the on-disk cache
   --cache-ttl   how long a cached page counts as fresh (default 15m)
   --data-dir    override the data directory
   --db          tee every record into a store
   --profile     load a named profile
   --dry-run     print what would happen
-q --quiet       suppress progress output
-v --verbose     more detail, repeatable
   --version     print the version

--tier takes a number or a surface name. --tier 0 refuses to use a session you have imported, which is how you check what an anonymous reader sees. --tier comet, graphql, og, embed or session pins one surface and fails rather than falling back, which is how you find out which surface a field actually came from.