diff --git a/DESCRIPTION b/DESCRIPTION
index 02147d3..b7ad052 100644
--- a/DESCRIPTION
+++ b/DESCRIPTION
@@ -1,8 +1,8 @@
 Package: chat.api
 Type: Package
 Title: Transport-Agnostic Chat Contract for R Agents
-Version: 0.0.1.27
-Date: 2026-09-07
+Version: 0.0.1.28
+Date: 2026-09-08
 Authors@R: c(
     person("Troy", "Hernandez", role = c("aut", "cre"),
            email = "troy@cornball.ai",
diff --git a/NEWS.md b/NEWS.md
index 12596a2..a912617 100644
--- a/NEWS.md
+++ b/NEWS.md
@@ -1,3 +1,14 @@
+# chat.api 0.0.1.28
+
+* Slack gains user-token identity: chat_slack(user_token =) plus
+  as_user = TRUE on chat_send() and chat_whoami() authenticate as an
+  actual workspace member instead of the bot, so a send shows up under
+  that member's real name and photo rather than the bot's profile.
+  Distinct from identity/username, which only relabels the bot's own
+  post. Capability flag user_identity reports whether a given Slack
+  client was configured with a user token; every other adapter reports
+  FALSE.
+
 # chat.api 0.0.1.27
 
 * New Telegram adapter, chat_telegram(), over the Bot API with HTTP
diff --git a/R/contract.R b/R/contract.R
index 96ad60f..ac06bc3 100644
--- a/R/contract.R
+++ b/R/contract.R
@@ -126,8 +126,12 @@ chat_resolve <- function(client, name, ...) {
 #'   media comes back out of \code{\link{chat_poll}} as
 #'   \code{\link{chat_attachment}} records), \code{typing},
 #'   \code{e2ee}, \code{identity_override} (logicals),
-#'   \code{markup_dialects} (character), \code{max_message_bytes}
-#'   (integer or NA).
+#'   \code{user_identity} (a send can authenticate as a real member of
+#'   the platform rather than as the bot -- Slack's
+#'   \code{chat_send(as_user = TRUE)} with a user token; a property of
+#'   the client instance's configuration there, and FALSE everywhere
+#'   else), \code{markup_dialects} (character),
+#'   \code{max_message_bytes} (integer or NA).
 #'
 #'   Sending and receiving get separate flags wherever a platform does
 #'   one and not the other, which is why \code{threads} and
diff --git a/R/irc.R b/R/irc.R
index 9b495b5..c4a2fce 100644
--- a/R/irc.R
+++ b/R/irc.R
@@ -133,7 +133,8 @@ chat_capabilities.chat_irc <- function(client, ...) {
          # has no join verb yet, so neither flag can be TRUE honestly.
          channel_create = FALSE, leave = FALSE, set_state = FALSE,
          files = FALSE, attachments = FALSE, typing = FALSE, e2ee = FALSE,
-         identity_override = FALSE, rich_markup = character(),
+         identity_override = FALSE, user_identity = FALSE,
+         rich_markup = character(),
          markup_dialects = "plain", max_message_bytes = 400L)
 }
 
diff --git a/R/loopback.R b/R/loopback.R
index 496c065..33fc55f 100644
--- a/R/loopback.R
+++ b/R/loopback.R
@@ -151,7 +151,8 @@ chat_capabilities.chat_loopback <- function(client, ...) {
          # them back out of the poll. Both TRUE is what makes loopback
          # the round-trip test double for media-carrying consumers.
          files = TRUE, attachments = TRUE, typing = FALSE, e2ee = FALSE,
-         identity_override = TRUE, rich_markup = character(),
+         identity_override = TRUE, user_identity = FALSE,
+         rich_markup = character(),
          markup_dialects = c("plain", "markdown"),
          max_message_bytes = NA_integer_)
 }
diff --git a/R/matrix.R b/R/matrix.R
index ad6b400..18ec908 100644
--- a/R/matrix.R
+++ b/R/matrix.R
@@ -972,7 +972,7 @@ chat_capabilities.chat_matrix <- function(client, ...) {
          # unmentioned.
          attachments = matrix_media_capable(client),
          typing = TRUE, e2ee = isTRUE(client$e2ee),
-         identity_override = FALSE,
+         identity_override = FALSE, user_identity = FALSE,
          # Empty on an e2ee client: the Megolm path builds its own HTML
          # from markdown and has nowhere to put a supplied fragment.
          rich_markup = if (isTRUE(client$e2ee)) character() else "html",
diff --git a/R/slack.R b/R/slack.R
index e92df50..3fe1e8d 100644
--- a/R/slack.R
+++ b/R/slack.R
@@ -27,9 +27,30 @@
 #' \code{chat_send(identity =)} (both need the chat:write.customize
 #' scope).
 #'
+#' \code{user_token} is a different kind of authorship than
+#' \code{identity}: \code{identity}/\code{username} relabel a bot's own
+#' post with a cosmetic name and icon, while a user token
+#' (\code{xoxp-...}, obtained via a Slack app's User Token Scopes
+#' rather than its Bot Token Scopes) authenticates as an actual
+#' workspace member, so \code{chat_send(..., as_user = TRUE)} and
+#' \code{chat_whoami(..., as_user = TRUE)} post and resolve identity as
+#' that member -- Slack shows their real name and photo, not a bot
+#' profile. Optional: leave unset if you only ever post as the bot.
+#'
+#' A post made \code{as_user = TRUE} is the member's own message in
+#' every respect, including to this client's own \code{\link{chat_poll}}:
+#' Slack messages carry no \code{self}, so nothing distinguishes it from
+#' something the member typed, and a consumer that replies to the
+#' member's traffic will reply to it.
+#'
 #' @param channels Character vector of channels to poll.
 #' @param token Bot token; defaults to the \code{SLACK_TOKEN}
 #'   environment variable.
+#' @param user_token User token (\code{xoxp-...}) for posting and
+#'   resolving identity as a real workspace member rather than the bot;
+#'   defaults to the \code{SLACK_USER_TOKEN} environment variable.
+#'   Optional -- leave unset (empty string) if \code{as_user} is never
+#'   used.
 #' @param username Default display-name override for sends, or NULL
 #'   (default) to post as the bot's own identity.
 #' @param .history Testing seam: replacement for
@@ -47,7 +68,9 @@
 #' @return A \code{chat_client} of class \code{chat_slack}.
 #' @export
 chat_slack <- function(channels = character(),
-                       token = Sys.getenv("SLACK_TOKEN"), username = NULL,
+                       token = Sys.getenv("SLACK_TOKEN"),
+                       user_token = Sys.getenv("SLACK_USER_TOKEN"),
+                       username = NULL,
                        .history = NULL, .post = NULL, .react = NULL,
                        .api = NULL) {
     if ((is.null(.history) || is.null(.post)) &&
@@ -61,7 +84,7 @@ chat_slack <- function(channels = character(),
     env <- new.env(parent = emptyenv())
     env$cursor <- list()
     structure(list(env = env, channels = sub("^#", "", channels),
-                   token = token, username = username,
+                   token = token, user_token = user_token, username = username,
                    history_fn = .history %||% slackr::slackr_history,
                    post_fn = .post %||% slackr::slackr_msg,
                    react_fn = .react, api_fn = .api),
@@ -176,25 +199,43 @@ chat_send.chat_slack <- function(client, channel, text,
                                  thread = NULL, reply_to = NULL,
                                  identity = NULL, files = NULL,
                                  kind = "message", notify = TRUE,
-                                 rich = NULL, ...) {
+                                 rich = NULL, as_user = FALSE, ...) {
     markup <- match.arg(markup)
-    name_override <- if (is.null(identity$name)) {
-        client$username
-    } else {
-        identity$name
+    if (isTRUE(as_user) && !nzchar(client$user_token %||% "")) {
+        stop("chat_send(as_user = TRUE) needs a user token ",
+             "(SLACK_USER_TOKEN).", call. = FALSE)
     }
-    # Empty strings suppress slackr's SLACK_USERNAME/SLACK_ICON_EMOJI
-    # env defaults (the same value an unset variable produces), so an
-    # ordinary send never silently inherits a customized identity
-    args <- list(txt = slack_render(text, markup),
-                 channel = sub("^#", "", channel),
-                 token = client$token,
-                 username = name_override %||% "",
-                 icon_emoji = if (is.null(identity$icon)) {
-            ""
+    # A user-token send authenticates as an actual member, so there is
+    # no bot identity left to relabel -- username/icon_emoji are a
+    # bot-only concept (chat:write.customize). They still go out as
+    # empty strings, for the same reason the bot path sends them:
+    # omitted, slackr fills them from SLACK_USERNAME/SLACK_ICON_EMOJI,
+    # and whether Slack ignores those on a user token is not something
+    # this adapter should have to know.
+    args <- if (isTRUE(as_user)) {
+        list(txt = slack_render(text, markup),
+             channel = sub("^#", "", channel),
+             token = client$user_token,
+             username = "", icon_emoji = "")
+    } else {
+        name_override <- if (is.null(identity$name)) {
+            client$username
         } else {
-            identity$icon
-        })
+            identity$name
+        }
+        # Empty strings suppress slackr's SLACK_USERNAME/SLACK_ICON_EMOJI
+        # env defaults (the same value an unset variable produces), so an
+        # ordinary send never silently inherits a customized identity
+        list(txt = slack_render(text, markup),
+             channel = sub("^#", "", channel),
+             token = client$token,
+             username = name_override %||% "",
+             icon_emoji = if (is.null(identity$icon)) {
+                 ""
+             } else {
+                 identity$icon
+             })
+    }
     if (identical(markup, "plain")) {
         # Rides slackr_msg's ... into the chat.postMessage body:
         # without it Slack styles *emphasis* even in "plain" text
@@ -419,24 +460,43 @@ chat_capabilities.chat_slack <- function(client, ...) {
          mark_read = TRUE, set_identity = TRUE, relogin = FALSE,
          channel_create = TRUE, leave = TRUE, set_state = FALSE,
          files = FALSE, attachments = FALSE, typing = FALSE, e2ee = FALSE,
-         identity_override = TRUE, rich_markup = character(),
+         identity_override = TRUE,
+         # Unlike every other flag here, this one is a property of the
+         # instance, not the adapter: whether as_user = TRUE will work
+         # depends on whether this client was built with a user_token,
+         # the same way a missing bot token already fails at
+         # construction rather than through a capability flag. It is
+         # surfaced here anyway because there is no construction-time
+         # failure to fail with -- a client is free to never use
+         # as_user, and unconfigured is a normal, valid state for it.
+         user_identity = nzchar(client$user_token %||% ""),
+         rich_markup = character(),
          markup_dialects = c("plain", "markdown"),
          max_message_bytes = 40000L)
 }
 
 #' @export
-chat_whoami.chat_slack <- function(client, ...) {
+chat_whoami.chat_slack <- function(client, as_user = FALSE, ...) {
     # Cached for the client's lifetime. Unlike Matrix, where the id is
     # already a field of the config in hand, Slack only answers this over
     # the network -- and chat_addressed() asks once per message. The
     # answer is a property of the token, which cannot change underneath a
     # client that was constructed with it.
-    if (!is.null(client$env$whoami)) {
-        return(client$env$whoami)
+    #
+    # A client can hold two identities now (bot token, user token), so
+    # each gets its own cache slot rather than sharing one.
+    cache_slot <- if (isTRUE(as_user)) "whoami_user" else "whoami"
+    if (!is.null(client$env[[cache_slot]])) {
+        return(client$env[[cache_slot]])
+    }
+    token <- if (isTRUE(as_user)) client$user_token else client$token
+    if (isTRUE(as_user) && !nzchar(token %||% "")) {
+        stop("chat_whoami(as_user = TRUE) needs a user token ",
+             "(SLACK_USER_TOKEN).", call. = FALSE)
     }
     api <- client$api_fn %||% slackr::call_slack_api
     body <- slack_body(api("/api/auth.test", .method = "GET",
-                           token = client$token))
+                           token = token))
     slack_stop_for_error(body, "auth.test")
     id <- body$user_id
     if (is.null(id) || !length(id) || !nzchar(as.character(id)[[1L]])) {
@@ -445,7 +505,7 @@ chat_whoami.chat_slack <- function(client, ...) {
     who <- chat_identity(as.character(id)[[1L]],
                          display = body$user %||% NA_character_,
                          raw = body)
-    client$env$whoami <- who
+    client$env[[cache_slot]] <- who
     who
 }
 
diff --git a/R/telegram.R b/R/telegram.R
index 43d88d4..39203b5 100644
--- a/R/telegram.R
+++ b/R/telegram.R
@@ -573,7 +573,8 @@ chat_capabilities.chat_telegram <- function(client, ...) {
          mark_read = FALSE, set_identity = TRUE, relogin = FALSE,
          channel_create = FALSE, leave = TRUE, set_state = FALSE,
          files = TRUE, attachments = TRUE, typing = TRUE, e2ee = FALSE,
-         identity_override = FALSE, rich_markup = "html",
+         identity_override = FALSE, user_identity = FALSE,
+         rich_markup = "html",
          markup_dialects = c("plain", "markdown"),
          # 4096 characters after entities parsing. Characters, not
          # bytes, so a message of multibyte text hits it sooner than
diff --git a/inst/tinytest/test_contract.R b/inst/tinytest/test_contract.R
index 5478ed1..a88f3ff 100644
--- a/inst/tinytest/test_contract.R
+++ b/inst/tinytest/test_contract.R
@@ -457,6 +457,23 @@ local({
     }
 })
 
+# ---- Every adapter answers user_identity ----
+# Slack's is a property of the client instance (was it built with a
+# user token), which is why it is worth checking that the others say
+# FALSE rather than nothing: a consumer that reads NULL cannot tell
+# "this adapter has no such thing" from "this client was not given
+# one". The stub client here has no user_token, so Slack answers FALSE
+# too; its TRUE case is in test_slack.R.
+local({
+    for (adapter in c("chat_loopback", "chat_irc", "chat_slack",
+                      "chat_matrix", "chat_telegram")) {
+        m <- getS3method("chat_capabilities", adapter)
+        caps <- m(structure(list(env = new.env()), class = adapter))
+        expect_true("user_identity" %in% names(caps), info = adapter)
+        expect_false(caps$user_identity, info = adapter)
+    }
+})
+
 # ---- Attachment record ----
 att <- chat_attachment("mxc://ex/abc", name = "plot.png",
                        mime = "image/png", bytes = 1024L)
diff --git a/inst/tinytest/test_slack.R b/inst/tinytest/test_slack.R
index 17c3347..ff2ebbf 100644
--- a/inst/tinytest/test_slack.R
+++ b/inst/tinytest/test_slack.R
@@ -519,6 +519,94 @@ local({
 
 expect_true(chat_capabilities(slack_api_client(function(...) NULL))$whoami)
 
+# ---- User-token identity ----
+# as_user authenticates as an actual workspace member (a user token)
+# rather than relabeling the bot's own post (identity/username, which
+# stays a cosmetic override on the bot's send). The two are different
+# mechanisms and get separate tests.
+
+local({
+    cl <- chat_slack(channels = "lab", token = "xoxb-bot", user_token = "xoxp-user",
+                     .history = function(...) NULL, .post = function(...) "1")
+    expect_identical(cl$user_token, "xoxp-user")
+})
+
+# A user-token send authenticates as the member, so there is no bot
+# identity left to relabel: username/icon_emoji are not sent.
+local({
+    seen <- NULL
+    fake_post <- function(...) {
+        seen <<- list(...)
+        list(ok = TRUE, ts = "999.1")
+    }
+    cl <- chat_slack(channels = "lab", token = "xoxb-bot", user_token = "xoxp-user",
+                     .history = function(...) NULL, .post = fake_post)
+    ts <- chat_send(cl, "lab", "as me", as_user = TRUE, thread = "42.1")
+    expect_identical(ts, "999.1")
+    expect_identical(seen$token, "xoxp-user")
+    # Sent as empty strings rather than omitted: omitted, slackr fills
+    # them from SLACK_USERNAME/SLACK_ICON_EMOJI, exactly what the bot
+    # path suppresses the same way.
+    expect_identical(seen$username, "")
+    expect_identical(seen$icon_emoji, "")
+    # Threads ride the same parameter on either token.
+    expect_identical(seen$thread_ts, "42.1")
+})
+
+# Asking to post as a member this client was never given a user token
+# for is a configuration error, not a silent fall-back to the bot.
+local({
+    cl <- chat_slack(channels = "lab", token = "xoxb-bot",
+                     .history = function(...) NULL,
+                     .post = function(...) list(ok = TRUE, ts = "1"))
+    expect_error(chat_send(cl, "lab", "as me", as_user = TRUE), "user token")
+})
+
+# chat_whoami(as_user = TRUE) resolves the member's identity, cached
+# separately from the bot's -- a client can legitimately hold both.
+local({
+    calls <- 0L
+    api <- function(path, ..., .method, token) {
+        calls <<- calls + 1L
+        if (identical(token, "xoxp-user")) {
+            list(ok = TRUE, user_id = "U0HUMAN", user = "jorge")
+        } else {
+            list(ok = TRUE, user_id = "U0BOT", user = "little-j")
+        }
+    }
+    cl <- chat_slack(channels = "lab", token = "xoxb-bot", user_token = "xoxp-user",
+                     .history = function(...) NULL, .post = function(...) "1", .api = api)
+    bot <- chat_whoami(cl)
+    me <- chat_whoami(cl, as_user = TRUE)
+    expect_identical(bot$id, "U0BOT")
+    expect_identical(me$id, "U0HUMAN")
+    expect_identical(calls, 2L)
+    # Each identity is cached on its own slot: neither call repeats.
+    chat_whoami(cl)
+    chat_whoami(cl, as_user = TRUE)
+    expect_identical(calls, 2L)
+})
+
+# chat_whoami(as_user = TRUE) on a client with no user token is the
+# same configuration error as the send path.
+local({
+    cl <- chat_slack(channels = "lab", token = "xoxb-bot",
+                     .history = function(...) NULL, .post = function(...) "1",
+                     .api = function(...) list(ok = TRUE, user_id = "U0BOT"))
+    expect_error(chat_whoami(cl, as_user = TRUE), "user token")
+})
+
+# The capability is a property of this instance's configuration, not a
+# static fact about the Slack adapter -- unlike every other flag here.
+local({
+    with_ut <- chat_slack(channels = "lab", token = "t", user_token = "xoxp-x",
+                          .history = function(...) NULL, .post = function(...) "1")
+    without_ut <- chat_slack(channels = "lab", token = "t", user_token = "",
+                             .history = function(...) NULL, .post = function(...) "1")
+    expect_true(chat_capabilities(with_ut)$user_identity)
+    expect_false(chat_capabilities(without_ut)$user_identity)
+})
+
 # ---- History paging ----
 local({
     seen <- NULL
diff --git a/man/chat_capabilities.Rd b/man/chat_capabilities.Rd
index 310c706..9ffc396 100644
--- a/man/chat_capabilities.Rd
+++ b/man/chat_capabilities.Rd
@@ -31,8 +31,12 @@ A list with at least: \code{threads} (can post into
   media comes back out of \code{\link{chat_poll}} as
   \code{\link{chat_attachment}} records), \code{typing},
   \code{e2ee}, \code{identity_override} (logicals),
-  \code{markup_dialects} (character), \code{max_message_bytes}
-  (integer or NA).
+  \code{user_identity} (a send can authenticate as a real member of
+  the platform rather than as the bot -- Slack's
+  \code{chat_send(as_user = TRUE)} with a user token; a property of
+  the client instance's configuration there, and FALSE everywhere
+  else), \code{markup_dialects} (character),
+  \code{max_message_bytes} (integer or NA).
 
   Sending and receiving get separate flags wherever a platform does
   one and not the other, which is why \code{threads} and
diff --git a/man/chat_slack.Rd b/man/chat_slack.Rd
index a74ca34..e0c51bb 100644
--- a/man/chat_slack.Rd
+++ b/man/chat_slack.Rd
@@ -6,6 +6,7 @@
 chat_slack(
   channels = character(),
   token = Sys.getenv("SLACK_TOKEN"),
+  user_token = Sys.getenv("SLACK_USER_TOKEN"),
   username = NULL,
   .history = NULL,
   .post = NULL,
@@ -19,6 +20,12 @@ chat_slack(
 \item{token}{Bot token; defaults to the \code{SLACK_TOKEN}
 environment variable.}
 
+\item{user_token}{User token (\code{xoxp-...}) for posting and
+resolving identity as a real workspace member rather than the bot;
+defaults to the \code{SLACK_USER_TOKEN} environment variable.
+Optional -- leave unset (empty string) if \code{as_user} is never
+used.}
+
 \item{username}{Default display-name override for sends, or NULL
 (default) to post as the bot's own identity.}
 
@@ -54,4 +61,20 @@ is only overridden through \code{username} here or
 \code{chat_send(identity =)} (both need the chat:write.customize
 scope).
 
+\code{user_token} is a different kind of authorship than
+\code{identity}: \code{identity}/\code{username} relabel a bot's own
+post with a cosmetic name and icon, while a user token
+(\code{xoxp-...}, obtained via a Slack app's User Token Scopes
+rather than its Bot Token Scopes) authenticates as an actual
+workspace member, so \code{chat_send(..., as_user = TRUE)} and
+\code{chat_whoami(..., as_user = TRUE)} post and resolve identity as
+that member -- Slack shows their real name and photo, not a bot
+profile. Optional: leave unset if you only ever post as the bot.
+
+A post made \code{as_user = TRUE} is the member's own message in
+every respect, including to this client's own \code{\link{chat_poll}}:
+Slack messages carry no \code{self}, so nothing distinguishes it from
+something the member typed, and a consumer that replies to the
+member's traffic will reply to it.
+
 }