diff --git a/integrations/tracabot.json b/integrations/tracabot.json
new file mode 100644
index 0000000..5f7ae7f
--- /dev/null
+++ b/integrations/tracabot.json
@@ -0,0 +1,65 @@
+{
+  "$schema": "../schema/integration.schema.json",
+  "schemaVersion": "0.1.0",
+  "slug": "tracabot",
+  "name": "TRACaBot",
+  "description": "TRACaBot is an OpenClaw-compatible Telegram anti-scam agent that writes and queries DKG v10 Working and Shared Memory so fraud reports, reviews, and campaign signals can be reused across communities.",
+  "category": ["openclaw", "telegram", "security", "shared-memory", "working-memory", "autoresearch"],
+  "maintainer": {
+    "github": "@brxtrac",
+    "name": "brxtrac"
+  },
+  "repo": "https://github.com/brxtrac/tracabot",
+  "commit": "89fb5aa708b7dafe5d293860a9c8a15a97f66d68",
+  "license": "MIT",
+  "requiresDkgNodeVersion": ">=10.0.0",
+  "memoryLayers": ["WM", "SWM"],
+  "v10PrimitivesUsed": [
+    "UAL",
+    "KnowledgeAsset",
+    "KnowledgeCollection",
+    "ContextGraph",
+    "Assertion",
+    "Integration",
+    "Curator",
+    "Entity"
+  ],
+  "publicInterfacesUsed": ["http-api"],
+  "targetAgents": ["OpenClaw", "generic-CLI", "generic-HTTP"],
+  "install": {
+    "kind": "service",
+    "runtime": "npm-global",
+    "npmGlobal": {
+      "package": "tracabot",
+      "version": "1.0.0",
+      "binary": "tracabot"
+    },
+    "envRequired": [
+      "TELEGRAM_BOT_TOKEN",
+      "DKG_NODE_URL",
+      "TRACABOT_ADMINS"
+    ]
+  },
+  "security": {
+    "networkEgress": [
+      "api.telegram.org",
+      "operator-configured local DKG node URL, default http://127.0.0.1:9200",
+      "operator-configured local OpenClaw gateway, default http://127.0.0.1:18789",
+      "operator-configured OpenAI-compatible LLM endpoint only when TRACABOT_LLM_BASE_URL is explicitly set"
+    ],
+    "writeAuthority": [
+      "createContextGraph via local DKG/OpenClaw daemon",
+      "createAssertion/writeAssertion/promoteAssertion for DKG Working Memory to Shared Working Memory staging when adapter-supported",
+      "SHARE evidence-backed fraud reports, findings, restrictions, bans, appeals, reviews, campaigns, and selected high-confidence channel observations via local DKG/OpenClaw daemon",
+      "query Context Graph Shared Memory with includeSharedMemory=true for cross-community risk evidence",
+      "PUBLISH qualifying high-confidence event roots via local DKG/OpenClaw daemon"
+    ],
+    "credentialsHandled": ["TELEGRAM_BOT_TOKEN", "optional LLM API key"],
+    "notes": "TRACaBot keeps DKG auth and OpenClaw credentials in the operator's local environment and does not print them in group chat. It has no preinstall/postinstall scripts, no remote code loading, no internal DKG v10 package imports, no daemon patches, and no endorsement/voting or Conviction UX. Curator-authority SHARE/PUBLISH operations are explicit runtime writes through the configured local DKG/OpenClaw adapter. Public Telegram replies redact internal UALs, graph names, OpenClaw endpoints, and admin setup details."
+  },
+  "trustTier": "community",
+  "designBrief": "https://github.com/brxtrac/tracabot/blob/89fb5aa708b7dafe5d293860a9c8a15a97f66d68/docs/DESIGN_BRIEF.md",
+  "demo": "https://github.com/brxtrac/tracabot/blob/89fb5aa708b7dafe5d293860a9c8a15a97f66d68/docs/demo/tracabot-demo.mp4",
+  "promotionPath": "TRACaBot keeps weak reports, watches, digests, and one-off join challenge state in local working memory. Evidence-backed reports, fraud findings, restrictions, bans, appeals, reviews, campaign roots, and selected high-confidence channel observations are staged through DKG Working Memory assertions when available and promoted to Shared Working Memory, with compatibility fallback to direct SHARE. High-confidence accepted reports, fraud findings, executed bans, admin-reviewed decisions, and repeated campaign roots are shaped as structured event entities for targeted publication and later Verified Memory/context-oracle consumption.",
+  "fitNotes": "TRACaBot applies the LLM-Wiki/autoresearch pattern to Telegram fraud defense: communities and OpenClaw agents produce structured claims, evidence, counter-evidence, appeals, reviews, and campaign clusters that can be queried and corrected across deployments instead of remaining isolated chat moderation events."
+}