diff --git a/.env.example b/.env.example
index 09364df..03a3002 100644
--- a/.env.example
+++ b/.env.example
@@ -42,3 +42,27 @@ OPENBRAIN__AUTH__ENABLED=false
 
 # Logging
 RUST_LOG=info,openbrain_mcp=debug
+
+# Truth Engine (optional)
+# Enable the background truth scoring worker. When enabled, stored memories
+# are continuously evaluated for truthfulness using PLN deduction, ECAN
+# attention economy, and cross-referencing against related memories.
+OPENBRAIN__TRUTH__ENABLED=false
+# Seconds between scoring cycles (default: 300 = 5 minutes)
+OPENBRAIN__TRUTH__SCORING_INTERVAL_SECONDS=300
+# Number of memories to score per cycle (default: 50)
+OPENBRAIN__TRUTH__BATCH_SIZE=50
+# Seconds before a scored memory is re-evaluated (default: 86400 = 24 hours)
+OPENBRAIN__TRUTH__RESCORE_AFTER_SECONDS=86400
+# PLN base confidence for deduction chains (0.0–1.0, default: 0.85)
+OPENBRAIN__TRUTH__PLN_BASE_CONFIDENCE=0.85
+# ECAN STI decay rate per cycle (0.0–1.0, default: 0.95)
+OPENBRAIN__TRUTH__ECAN_DECAY_RATE=0.95
+# ECAN attention spread factor (default: 0.05)
+OPENBRAIN__TRUTH__ECAN_SPREAD_FACTOR=0.05
+# Similarity threshold for contradiction detection (0.0–1.0, default: 0.85)
+OPENBRAIN__TRUTH__CONTRADICTION_THRESHOLD=0.85
+# Truth value threshold for "verified" categorization (0.0–1.0, default: 0.8)
+OPENBRAIN__TRUTH__VERIFICATION_THRESHOLD=0.8
+# Max related memories to cross-reference per scoring (default: 10)
+OPENBRAIN__TRUTH__CROSS_REF_LIMIT=10
diff --git a/src/config.rs b/src/config.rs
index e9aa29f..c43c351 100644
--- a/src/config.rs
+++ b/src/config.rs
@@ -15,6 +15,7 @@ pub struct Config {
     pub dedup: DedupConfig,
     pub ttl: TtlConfig,
     pub auth: AuthConfig,
+    pub truth: TruthConfig,
 }
 
 /// Server configuration
@@ -71,6 +72,41 @@ pub struct TtlConfig {
     pub cleanup_interval_seconds: u64,
 }
 
+/// Truth scoring engine configuration
+#[derive(Debug, Clone, Deserialize)]
+pub struct TruthConfig {
+    /// Enable truth scoring background worker
+    #[serde(default = "default_truth_enabled")]
+    pub enabled: bool,
+    /// Seconds between scoring cycles
+    #[serde(default = "default_scoring_interval_seconds")]
+    pub scoring_interval_seconds: u64,
+    /// Number of memories to score per cycle
+    #[serde(default = "default_truth_batch_size")]
+    pub batch_size: usize,
+    /// Seconds before a scored memory is re-evaluated
+    #[serde(default = "default_rescore_after_seconds")]
+    pub rescore_after_seconds: u64,
+    /// Base confidence for PLN deduction chains
+    #[serde(default = "default_pln_base_confidence")]
+    pub pln_base_confidence: f32,
+    /// ECAN STI decay rate per cycle (0.0–1.0)
+    #[serde(default = "default_ecan_decay_rate")]
+    pub ecan_decay_rate: f32,
+    /// ECAN attention spread factor
+    #[serde(default = "default_ecan_spread_factor")]
+    pub ecan_spread_factor: f32,
+    /// Similarity threshold above which conflicting memories are contradictions
+    #[serde(default = "default_contradiction_threshold")]
+    pub contradiction_threshold: f32,
+    /// Truth value threshold for "verified" categorization
+    #[serde(default = "default_verification_threshold")]
+    pub verification_threshold: f32,
+    /// Max related memories to cross-reference per scoring
+    #[serde(default = "default_cross_ref_limit")]
+    pub cross_ref_limit: i64,
+}
+
 /// Authentication configuration
 #[derive(Debug, Clone, Deserialize)]
 pub struct AuthConfig {
@@ -139,6 +175,38 @@ fn default_auth_enabled() -> bool {
     false
 }
 
+// Truth engine defaults
+fn default_truth_enabled() -> bool {
+    false
+}
+fn default_scoring_interval_seconds() -> u64 {
+    300
+}
+fn default_truth_batch_size() -> usize {
+    50
+}
+fn default_rescore_after_seconds() -> u64 {
+    86400
+}
+fn default_pln_base_confidence() -> f32 {
+    0.85
+}
+fn default_ecan_decay_rate() -> f32 {
+    0.95
+}
+fn default_ecan_spread_factor() -> f32 {
+    0.05
+}
+fn default_contradiction_threshold() -> f32 {
+    0.85
+}
+fn default_verification_threshold() -> f32 {
+    0.8
+}
+fn default_cross_ref_limit() -> i64 {
+    10
+}
+
 impl Config {
     /// Load configuration from environment variables
     pub fn load() -> Result<Self> {
@@ -167,6 +235,41 @@ impl Config {
             )?
             // Auth settings
             .set_default("auth.enabled", default_auth_enabled())?
+            // Truth engine settings
+            .set_default("truth.enabled", default_truth_enabled())?
+            .set_default(
+                "truth.scoring_interval_seconds",
+                default_scoring_interval_seconds() as i64,
+            )?
+            .set_default("truth.batch_size", default_truth_batch_size() as i64)?
+            .set_default(
+                "truth.rescore_after_seconds",
+                default_rescore_after_seconds() as i64,
+            )?
+            .set_default(
+                "truth.pln_base_confidence",
+                default_pln_base_confidence() as f64,
+            )?
+            .set_default(
+                "truth.ecan_decay_rate",
+                default_ecan_decay_rate() as f64,
+            )?
+            .set_default(
+                "truth.ecan_spread_factor",
+                default_ecan_spread_factor() as f64,
+            )?
+            .set_default(
+                "truth.contradiction_threshold",
+                default_contradiction_threshold() as f64,
+            )?
+            .set_default(
+                "truth.verification_threshold",
+                default_verification_threshold() as f64,
+            )?
+            .set_default(
+                "truth.cross_ref_limit",
+                default_cross_ref_limit(),
+            )?
             // Load from environment with OPENBRAIN_ prefix
             .add_source(
                 config::Environment::with_prefix("OPENBRAIN")
@@ -231,6 +334,17 @@ impl Default for Config {
                 enabled: default_auth_enabled(),
                 api_keys: Vec::new(),
             },
-        }
+            truth: TruthConfig {
+                enabled: default_truth_enabled(),
+                scoring_interval_seconds: default_scoring_interval_seconds(),
+                batch_size: default_truth_batch_size(),
+                rescore_after_seconds: default_rescore_after_seconds(),
+                pln_base_confidence: default_pln_base_confidence(),
+                ecan_decay_rate: default_ecan_decay_rate(),
+                ecan_spread_factor: default_ecan_spread_factor(),
+                contradiction_threshold: default_contradiction_threshold(),
+                verification_threshold: default_verification_threshold(),
+                cross_ref_limit: default_cross_ref_limit(),
+            },
     }
 }