diff --git a/docker-compose.yaml b/docker-compose.yaml
index af9261f..41cd688 100644
--- a/docker-compose.yaml
+++ b/docker-compose.yaml
@@ -28,9 +28,51 @@ services:
       # GitHub MCP (Remote)
       - GITHUB_PERSONAL_ACCESS_TOKEN=${GITHUB_PERSONAL_ACCESS_TOKEN}
       - GITHUB_MCP_URL=https://api.githubcopilot.com/mcp/
+      # LiteLLM
+      - LITELLM_URL=http://litellm:4000
+      - LITELLM_MASTER_KEY=sk-sre-agent-local
     depends_on:
       slack:
         condition: service_started
+      litellm:
+        condition: service_healthy
+    restart: unless-stopped
+
+  # LiteLLM Proxy (cost tracking + model routing)
+  litellm:
+    image: ghcr.io/berriai/litellm:main-latest
+    command: ["--config", "/app/litellm_config.yaml"]
+    volumes:
+      - ./litellm_config.yaml:/app/litellm_config.yaml:ro
+    environment:
+      - ANTHROPIC_API_KEY=${ANTHROPIC_API_KEY}
+      - DATABASE_URL=postgresql://litellm:litellm@postgres:5432/litellm
+    ports:
+      - "4000:4000"
+    healthcheck:
+      test: ["CMD-SHELL", "curl -f http://localhost:4000/health/liveliness || exit 1"]
+      interval: 10s
+      timeout: 5s
+      retries: 5
+    depends_on:
+      postgres:
+        condition: service_healthy
+    restart: unless-stopped
+
+  # PostgreSQL (for LiteLLM cost tracking)
+  postgres:
+    image: postgres:16
+    environment:
+      - POSTGRES_USER=litellm
+      - POSTGRES_PASSWORD=litellm
+      - POSTGRES_DB=litellm
+    volumes:
+      - postgres_data:/var/lib/postgresql/data
+    healthcheck:
+      test: ["CMD-SHELL", "pg_isready -U litellm -d litellm"]
+      interval: 5s
+      timeout: 5s
+      retries: 5
     restart: unless-stopped
 
   # Slack MCP Server (korotovsky/slack-mcp-server)
@@ -44,3 +86,6 @@ services:
     restart: unless-stopped
     ports:
       - "13080:13080"
+
+volumes:
+  postgres_data:
diff --git a/litellm_config.yaml b/litellm_config.yaml
new file mode 100644
index 0000000..ca2bead
--- /dev/null
+++ b/litellm_config.yaml
@@ -0,0 +1,9 @@
+model_list:
+  - model_name: claude-sonnet-4-5
+    litellm_params:
+      model: anthropic/claude-sonnet-4-5-20250929
+      api_key: os.environ/ANTHROPIC_API_KEY
+
+general_settings:
+  master_key: sk-sre-agent-local
+  database_url: "postgresql://litellm:litellm@postgres:5432/litellm"
diff --git a/src/sre_agent/core/agent.py b/src/sre_agent/core/agent.py
index caa703c..473a1cb 100644
--- a/src/sre_agent/core/agent.py
+++ b/src/sre_agent/core/agent.py
@@ -1,6 +1,8 @@
 """SRE Agent using pydantic-ai."""
 
 from pydantic_ai import Agent
+from pydantic_ai.models.anthropic import AnthropicModel
+from pydantic_ai.providers.anthropic import AnthropicProvider
 
 from sre_agent.core.models import ErrorDiagnosis
 from sre_agent.core.prompts import SYSTEM_PROMPT, build_diagnosis_prompt
@@ -27,8 +29,16 @@ def create_sre_agent(config: AgentSettings) -> Agent[None, ErrorDiagnosis]:
         create_slack_mcp_toolset(config),
     ]
 
-    return Agent(
+    model = AnthropicModel(
         config.model,
+        provider=AnthropicProvider(
+            api_key=config.litellm.master_key,
+            base_url=f"{config.litellm.url}/anthropic/",
+        ),
+    )
+
+    return Agent(
+        model,
         system_prompt=SYSTEM_PROMPT,
         output_type=ErrorDiagnosis,
         toolsets=toolsets,
diff --git a/src/sre_agent/core/settings.py b/src/sre_agent/core/settings.py
index d2d2cfc..cc610a9 100644
--- a/src/sre_agent/core/settings.py
+++ b/src/sre_agent/core/settings.py
@@ -50,6 +50,19 @@ class SlackSettings(BaseSettings):
     mcp_url: str = Field(description="URL of Slack MCP server (SSE)")
 
 
+class LiteLLMSettings(BaseSettings):
+    """LiteLLM proxy configuration."""
+
+    model_config = SettingsConfigDict(
+        env_prefix="LITELLM_",
+        env_file=ENV_FILE_PATH,
+        extra="ignore",
+    )
+
+    url: str = Field(default="http://litellm:4000", description="LiteLLM proxy URL")
+    master_key: str = Field(description="LiteLLM master key")
+
+
 class AgentSettings(BaseSettings):
     """Main agent configuration."""
 
@@ -67,6 +80,7 @@ class AgentSettings(BaseSettings):
     aws: AWSSettings
     github: GitHubSettings
     slack: SlackSettings
+    litellm: LiteLLMSettings
 
 
 def get_settings() -> AgentSettings:
@@ -84,4 +98,5 @@ def get_settings() -> AgentSettings:
         aws=AWSSettings(),
         github=GitHubSettings(),  # type: ignore[call-arg]
         slack=SlackSettings(),  # type: ignore[call-arg]
+        litellm=LiteLLMSettings(),  # type: ignore[call-arg]
     )