diff --git a/README.md b/README.md
index 2cbbaf4..6bed23e 100644
--- a/README.md
+++ b/README.md
@@ -68,7 +68,7 @@ docker build . -t nerve
 
 ## LLM Support
 
-Nerve features integrations for any model accessible via the [ollama](https://github.com/ollama/ollama), [groq](https://groq.com), [OpenAI](https://openai.com/index/openai-api/), [Anthropic](https://www.anthropic.com/), [Fireworks](https://fireworks.ai/), [Huggingface](https://huggingface.co/blog/tgi-messages-api#using-inference-endpoints-with-openai-client-libraries), [Nvidia NIM](https://www.nvidia.com/en-us/ai/) and [NovitaAI](https://novita.ai/model-api/product/llm-api) APIs. 
+Nerve features integrations for any model accessible via the [ollama](https://github.com/ollama/ollama), [groq](https://groq.com), [OpenAI](https://openai.com/index/openai-api/), [Anthropic](https://www.anthropic.com/), [Fireworks](https://fireworks.ai/), [Huggingface](https://huggingface.co/blog/tgi-messages-api#using-inference-endpoints-with-openai-client-libraries), [Nvidia NIM](https://www.nvidia.com/en-us/ai/), [DeepSeek](https://deepseek.com/) and [NovitaAI](https://novita.ai/model-api/product/llm-api) APIs. 
 
 **The tool will automatically detect if the selected model natively supports function calling. If not, it will provide a compatibility layer that empowers older models to perform function calling anyway.**
 
@@ -118,6 +118,12 @@ For **Nvidia NIM**:
 NIM_API_KEY=you-api-key nerve -G "nim://nvidia/nemotron-4-340b-instruct" ...
 ```
 
+For **DeepSeek**:
+
+```sh
+DEEPSEEK_API_KEY=you-api-key nerve -G "deepseek://deepseek-chat" ...
+```
+
 For **Novita**:
 
 ```sh
diff --git a/src/agent/generator/deepseek.rs b/src/agent/generator/deepseek.rs
new file mode 100644
index 0000000..c07f44c
--- /dev/null
+++ b/src/agent/generator/deepseek.rs
@@ -0,0 +1,45 @@
+use anyhow::Result;
+use async_trait::async_trait;
+
+use crate::agent::state::SharedState;
+
+use super::{openai::OpenAIClient, ChatOptions, ChatResponse, Client};
+
+pub struct DeepSeekClient {
+    client: OpenAIClient,
+}
+
+#[async_trait]
+impl Client for DeepSeekClient {
+    fn new(_: &str, _: u16, model_name: &str, _: u32) -> anyhow::Result<Self>
+    where
+        Self: Sized,
+    {
+        let client = OpenAIClient::custom(
+            model_name,
+            "DEEPSEEK_API_KEY",
+            "https://api.deepseek.com/v1/",
+        )?;
+
+        Ok(Self { client })
+    }
+
+    async fn check_native_tools_support(&self) -> Result<bool> {
+        self.client.check_native_tools_support().await
+    }
+
+    async fn chat(
+        &self,
+        state: SharedState,
+        options: &ChatOptions,
+    ) -> anyhow::Result<ChatResponse> {
+        self.client.chat(state, options).await
+    }
+}
+
+#[async_trait]
+impl mini_rag::Embedder for DeepSeekClient {
+    async fn embed(&self, text: &str) -> Result<mini_rag::Embeddings> {
+        self.client.embed(text).await
+    }
+}
diff --git a/src/agent/generator/mod.rs b/src/agent/generator/mod.rs
index a05b7e0..6774f6c 100644
--- a/src/agent/generator/mod.rs
+++ b/src/agent/generator/mod.rs
@@ -10,6 +10,7 @@ use serde::{Deserialize, Serialize};
 use super::{state::SharedState, Invocation};
 
 mod anthropic;
+mod deepseek;
 mod fireworks;
 mod groq;
 mod huggingface;
@@ -194,6 +195,12 @@ macro_rules! factory_body {
                 $model_name,
                 $context_window,
             )?)),
+            "deepseek" => Ok(Box::new(deepseek::DeepSeekClient::new(
+                $url,
+                $port,
+                $model_name,
+                $context_window,
+            )?)),
             "http" => Ok(Box::new(openai_compatible::OpenAiCompatibleClient::new(
                 $url,
                 $port,