Channels baseline: email, Signal, Matrix, Mattermost, Teams, Google Chat, SMS + media and voice

cmd/agent_context.go

@@ -4,16 +4,53 @@ import (
 	"encoding/json"
 	"os"
 
+	"github.com/memcode-ai/memcode/internal/agent/input"
 	"github.com/memcode-ai/memcode/internal/agent/runtime"
+	"github.com/memcode-ai/memcode/internal/channels"
 	gwconfig "github.com/memcode-ai/memcode/internal/gateway/config"
 )
 
 // jobContext mirrors the envelope the gateway persists at gwconfig.ContextPath —
-// the persona's supplemental context plus its extra skill roots. The JSON shape
-// is the contract with internal/gateway/server.
+// the persona's supplemental context, its extra skill roots, and this task's
+// media as spool IDs. The JSON shape is the contract with internal/gateway/server.
 type jobContext struct {
 	Items      []runtime.ContextItem `json:"items,omitempty"`
 	SkillRoots []string              `json:"skill_roots,omitempty"`
+	// Attachments are media spool IDs — bare <sha256>.<ext> filenames, resolved
+	// STRICTLY inside the gateway media spool (see resolveJobAttachments). Never
+	// paths: the spool is the trust boundary, so a corrupted context file cannot
+	// point this job at arbitrary local files.
+	Attachments []string `json:"attachments,omitempty"`
+}
+
+// resolveJobAttachments turns spool IDs into engine attachments. Each ID must
+// resolve inside the media spool; anything else — separators, dot-files, a
+// missing file, an unsupported kind — is skipped. Audio never reaches the
+// engine (the gateway transcribes it before spawning the job).
+func resolveJobAttachments(ids []string) []input.Attachment {
+	if len(ids) == 0 {
+		return nil
+	}
+	spool, err := gwconfig.MediaDir()
+	if err != nil {
+		return nil
+	}
+	var out []input.Attachment
+	for _, id := range ids {
+		path, err := channels.ResolveSpoolID(spool, id)
+		if err != nil {
+			continue
+		}
+		att, ok := input.Resolve(path, spool, "channel")
+		if !ok {
+			continue
+		}
+		switch att.Kind {
+		case input.KindImage, input.KindPDF, input.KindText:
+			out = append(out, att)
+		}
+	}
+	return out
 }
 
 // loadJobContext reads the job context the gateway persisted for this session

cmd/gateway.go

@@ -99,7 +99,7 @@ var gatewaySetupCmd = &cobra.Command{
 			} else {
 				cmd.Printf("Configured: %s\n", strings.Join(enabled, ", "))
 			}
-			choice := strings.ToLower(strings.TrimSpace(prompt(in, cmd, "Channel to add/update [telegram/discord/slack/github/whatsapp] (blank to finish): ")))
+			choice := strings.ToLower(strings.TrimSpace(prompt(in, cmd, "Channel to add/update [telegram/discord/slack/email/signal/matrix/mattermost/msteams/googlechat/sms/github/whatsapp] (blank to finish): ")))
 
 			secrets := map[string]string{}
 			if settings.Channels == nil {
@@ -136,8 +136,46 @@ var gatewaySetupCmd = &cobra.Command{
 				secrets[gwconfig.EnvWhatsAppVerify] = secret(cmd, "Webhook verify token: ")
 				secrets[gwconfig.EnvWhatsAppSecret] = secret(cmd, "App secret (verifies inbound; required to activate): ")
 				ch.AllowFrom = allowList(in, cmd)
+			case "email":
+				cmd.Println("Use a DEDICATED mailbox (an app password for Gmail/Outlook), never your personal inbox.")
+				secrets[gwconfig.EnvEmailAddress] = strings.TrimSpace(prompt(in, cmd, "Email address: "))
+				secrets[gwconfig.EnvEmailPassword] = secret(cmd, "App password: ")
+				secrets[gwconfig.EnvEmailIMAPHost] = strings.TrimSpace(prompt(in, cmd, "IMAP host (e.g. imap.gmail.com): "))
+				secrets[gwconfig.EnvEmailSMTPHost] = strings.TrimSpace(prompt(in, cmd, "SMTP host (e.g. smtp.gmail.com): "))
+				ch.AllowFrom = allowList(in, cmd)
+			case "signal":
+				cmd.Println("Requires a running signal-cli daemon in HTTP mode (see the docs); use a dedicated number.")
+				secrets[gwconfig.EnvSignalNumber] = strings.TrimSpace(prompt(in, cmd, "Your Signal number (+E.164): "))
+				if u := strings.TrimSpace(prompt(in, cmd, "signal-cli daemon URL (blank = http://127.0.0.1:8080): ")); u != "" {
+					secrets[gwconfig.EnvSignalCLIURL] = u
+				}
+				ch.AllowFrom = allowList(in, cmd)
+			case "matrix":
+				cmd.Println("Plain rooms only for now (no end-to-end-encrypted rooms).")
+				secrets[gwconfig.EnvMatrixHomeserver] = strings.TrimSpace(prompt(in, cmd, "Homeserver URL (e.g. https://matrix.org): "))
+				secrets[gwconfig.EnvMatrixToken] = secret(cmd, "Access token: ")
+				ch.AllowFrom = allowList(in, cmd)
+			case "mattermost":
+				secrets[gwconfig.EnvMattermostURL] = strings.TrimSpace(prompt(in, cmd, "Server URL (e.g. https://mm.example.com): "))
+				secrets[gwconfig.EnvMattermostToken] = secret(cmd, "Bot access token: ")
+				ch.AllowFrom = allowList(in, cmd)
+			case "msteams":
+				secrets[gwconfig.EnvTeamsAppID] = strings.TrimSpace(prompt(in, cmd, "Azure app (bot) ID: "))
+				secrets[gwconfig.EnvTeamsAppPassword] = secret(cmd, "Client secret: ")
+				secrets[gwconfig.EnvTeamsTenantID] = strings.TrimSpace(prompt(in, cmd, "Tenant ID: "))
+				ch.AllowFrom = allowList(in, cmd)
+			case "googlechat":
+				secrets[gwconfig.EnvGoogleChatSAKey] = strings.TrimSpace(prompt(in, cmd, "Path to the service-account JSON key: "))
+				ch.Audience = strings.TrimSpace(prompt(in, cmd, "Project number (JWT audience): "))
+				ch.AllowFrom = allowList(in, cmd)
+			case "sms":
+				secrets[gwconfig.EnvTwilioAccountSID] = strings.TrimSpace(prompt(in, cmd, "Twilio Account SID: "))
+				secrets[gwconfig.EnvTwilioAuthToken] = secret(cmd, "Auth token: ")
+				secrets[gwconfig.EnvTwilioFromNumber] = strings.TrimSpace(prompt(in, cmd, "Your Twilio number (+E.164): "))
+				ch.WebhookURL = strings.TrimSpace(prompt(in, cmd, "Exact public webhook URL (e.g. https://gw.example.com/webhook/sms): "))
+				ch.AllowFrom = allowList(in, cmd)
 			default:
-				cmd.Println("Unknown channel; pick one of telegram/discord/slack/github/whatsapp.")
+				cmd.Println("Unknown channel; pick one of telegram/discord/slack/email/signal/matrix/mattermost/msteams/googlechat/sms/github/whatsapp.")
 				continue
 			}
 			settings.Channels[choice] = ch

cmd/run.go

@@ -141,9 +141,10 @@ for local gateway development. Never store keys in .memcode.`,
 			// Uses the chat seams (which load + save the transcript) instead of Run.
 			if sessionID, _ := cmd.Flags().GetString("session"); sessionID != "" {
 				sess.SetSessionID(sessionID)
-				if jc := loadJobContext(sessionID); len(jc.Items) > 0 || len(jc.SkillRoots) > 0 {
-					sess.SetContext(jc.Items)         // gateway-supplied persona/user context for this run
-					sess.SetSkillRoots(jc.SkillRoots) // persona's own skills join discovery
+				if jc := loadJobContext(sessionID); len(jc.Items) > 0 || len(jc.SkillRoots) > 0 || len(jc.Attachments) > 0 {
+					sess.SetContext(jc.Items)                                      // gateway-supplied persona/user context for this run
+					sess.SetSkillRoots(jc.SkillRoots)                              // persona's own skills join discovery
+					sess.SetTaskAttachments(resolveJobAttachments(jc.Attachments)) // channel media rides this turn
 				}
 				if _, err := runtime.ResolveSession(cfg.Root, sessionID); err == nil {
 					sess.SetResume(sessionID)

docs/gateway/README.md

@@ -2,8 +2,9 @@
 
 The same `memcode` binary that runs the interactive agent can run as a
 long-lived, self-hosted **gateway**: it listens on the surfaces people already
-use (Telegram, Discord, Slack, GitHub, WhatsApp), turns each inbound message
-into an agent job, and posts the result back. Coding is one use of this loop,
+use (Telegram, Discord, Slack, Email, Signal, Matrix, Mattermost, Microsoft
+Teams, Google Chat, SMS, GitHub, WhatsApp), turns each inbound message into an
+agent job, and posts the result back. Coding is one use of this loop,
 not what it's built around — an inbound message is just a task.
 
 ```
@@ -30,13 +31,31 @@ It routes each answer the way memcode splits configuration:
 
 A channel is enabled when its secret is present.
 
-| Channel  | Secret(s) in `.env`                                        | Transport         |
-|----------|------------------------------------------------------------|-------------------|
-| Telegram | `TELEGRAM_BOT_TOKEN`                                        | Bot API long-poll |
-| Discord  | `DISCORD_BOT_TOKEN`                                         | gateway websocket |
-| Slack    | `SLACK_APP_TOKEN`, `SLACK_BOT_TOKEN`                        | Socket Mode       |
-| GitHub   | `GITHUB_WEBHOOK_SECRET`                                     | inbound webhook   |
-| WhatsApp | `WHATSAPP_ACCESS_TOKEN`, `WHATSAPP_VERIFY_TOKEN`, `WHATSAPP_APP_SECRET` | Meta Cloud API |
+| Channel     | Secret(s) in `.env`                                        | Transport         |
+|-------------|------------------------------------------------------------|-------------------|
+| Telegram    | `TELEGRAM_BOT_TOKEN`                                        | Bot API long-poll |
+| Discord     | `DISCORD_BOT_TOKEN`                                         | gateway websocket |
+| Slack       | `SLACK_APP_TOKEN`, `SLACK_BOT_TOKEN`                        | Socket Mode       |
+| Email       | `EMAIL_ADDRESS`, `EMAIL_PASSWORD`, `EMAIL_IMAP_HOST`, `EMAIL_SMTP_HOST` | IMAP poll + SMTP |
+| Signal      | `SIGNAL_NUMBER` (+ optional `SIGNAL_CLI_URL`)               | signal-cli daemon (SSE + JSON-RPC) |
+| Matrix      | `MATRIX_HOMESERVER`, `MATRIX_ACCESS_TOKEN`                  | client-server /sync (no E2EE v1) |
+| Mattermost  | `MATTERMOST_URL`, `MATTERMOST_TOKEN`                        | websocket + REST v4 |
+| MS Teams    | `TEAMS_APP_ID`, `TEAMS_APP_PASSWORD`, `TEAMS_TENANT_ID`     | Bot Framework webhook |
+| Google Chat | `GOOGLE_CHAT_SA_KEY` (path) + `googlechat.audience`         | signed webhook + Chat REST |
+| SMS         | `TWILIO_ACCOUNT_SID`, `TWILIO_AUTH_TOKEN`, `TWILIO_FROM_NUMBER` + `sms.webhook_url` | Twilio webhook + Messages API |
+| GitHub      | `GITHUB_WEBHOOK_SECRET`                                     | inbound webhook   |
+| WhatsApp    | `WHATSAPP_ACCESS_TOKEN`, `WHATSAPP_VERIFY_TOKEN`, `WHATSAPP_APP_SECRET` | Meta Cloud API |
+
+Webhook-driven surfaces (Teams, Google Chat, SMS, GitHub, WhatsApp) mount on
+the shared listener (`webhook.addr`, default `:8787`) at
+`/webhook/{teams,googlechat,sms,github,whatsapp}` — expose it over HTTPS.
+Email dedup is keyed on `<mailbox>/<UIDVALIDITY>/<UID>` (the provider-side ack
+identity); Message-ID serves threading only. Email's sender identity is the
+RFC From address — weaker than the other channels' platform-verified ids, so
+its allow-list depends on your mailbox provider rejecting spoofed mail
+(SPF/DKIM/DMARC); use a mainstream provider and a dedicated account. Signal requires a signal-cli
+daemon in native HTTP mode; Matrix v1 is plain rooms only (E2EE is a known
+follow-up).
 
 ### gateway.yaml
 
@@ -136,6 +155,23 @@ projects, agents, channel knobs) on change, so an approval takes effect within
 seconds — no restart. Channel connections and schedules are wired at startup and
 do not hot-reload.
 
+## Media and voice
+
+Inbound attachments (photos, PDFs, documents) are downloaded into a
+content-addressed media spool (`~/.config/memcode/media`, pruned with the
+inbox) and ride the task into the engine as native image/document blocks.
+Everything downstream of the adapter addresses media by spool ID, never by
+path — the spool is the trust boundary.
+
+Voice notes are transcribed gateway-side (OpenAI `gpt-4o-mini-transcribe`
+falling back to `whisper-1`, or Gemini — picked by whichever key is present)
+and the transcript becomes the task text; audio never reaches the engine.
+Without either key a voice-only message gets an honest "not configured" reply.
+Optionally, `channels.<name>.voice_replies: in_kind|always` (default `off`)
+synthesizes an OGG/Opus voice reply (OpenAI `gpt-4o-mini-tts` — the full text
+is always sent alongside, code blocks are never spoken, synthesis failures
+degrade to text).
+
 ## Import from OpenClaw
 
 Already running OpenClaw? Bring your channels over with one command:

go.mod

@@ -14,6 +14,9 @@ require (
 	github.com/charmbracelet/x/term v0.2.2
 	github.com/chromedp/cdproto v0.0.0-20260321001828-e3e3800016bc
 	github.com/chromedp/chromedp v0.15.1
+	github.com/emersion/go-imap/v2 v2.0.0-beta.8
+	github.com/golang-jwt/jwt/v5 v5.3.1
+	github.com/gorilla/websocket v1.5.3
 	github.com/mattn/go-runewidth v0.0.24
 	github.com/modelcontextprotocol/go-sdk v1.6.1
 	github.com/openai/openai-go/v3 v3.41.1
@@ -24,6 +27,7 @@ require (
 	go.yaml.in/yaml/v4 v4.0.0-rc.2
 	golang.org/x/image v0.43.0
 	golang.org/x/net v0.56.0
+	golang.org/x/oauth2 v0.36.0
 	golang.org/x/sys v0.46.0
 	golang.org/x/term v0.44.0
 	google.golang.org/genai v1.63.0
@@ -45,6 +49,8 @@ require (
 	github.com/clipperhouse/uax29/v2 v2.7.0 // indirect
 	github.com/dlclark/regexp2/v2 v2.1.1 // indirect
 	github.com/dustin/go-humanize v1.0.1 // indirect
+	github.com/emersion/go-message v0.18.2 // indirect
+	github.com/emersion/go-sasl v0.0.0-20241020182733-b788ff22d5a6 // indirect
 	github.com/felixge/httpsnoop v1.0.4 // indirect
 	github.com/go-json-experiment/json v0.0.0-20260214004413-d219187c3433 // indirect
 	github.com/go-logr/logr v1.4.3 // indirect
@@ -58,7 +64,6 @@ require (
 	github.com/google/uuid v1.6.0 // indirect
 	github.com/googleapis/enterprise-certificate-proxy v0.3.17 // indirect
 	github.com/googleapis/gax-go/v2 v2.23.0 // indirect
-	github.com/gorilla/websocket v1.5.3 // indirect
 	github.com/inconshreveable/mousetrap v1.1.0 // indirect
 	github.com/invopop/jsonschema v0.14.0 // indirect
 	github.com/lucasb-eyer/go-colorful v1.4.0 // indirect
@@ -84,7 +89,6 @@ require (
 	go.opentelemetry.io/otel/metric v1.44.0 // indirect
 	go.opentelemetry.io/otel/trace v1.44.0 // indirect
 	golang.org/x/crypto v0.53.0 // indirect
-	golang.org/x/oauth2 v0.36.0 // indirect
 	golang.org/x/sync v0.21.0 // indirect
 	golang.org/x/text v0.38.0 // indirect
 	google.golang.org/api v0.287.1 // indirect