DeepSeek Harness to Claude code integration
Budget / SalaryHourly project
TypeFreelance project
LocationRemote
Posted1 hour ago
ALL WORK DONE THROUGH SCREENSHARE
### DeepSeek Harness (DSH) & Claude Code Integration Expert (with Memory plugin Setup)
**Project Overview**
I need a developer to configure a meta-agent workspace integrating DeepSeek Harness (DSH), Claude Code, and Mem0. The goal is to use DSH as the workspace orchestrator to handle prompt caching and file execution, route the actual reasoning through Claude Code, and use Mem0 to retain long-term project memory across sessions.
**Core Technical Requirements**
* **Architecture & Routing:** Deploy DeepSeek Harness (`npx @deepseek-ai/dsh web`) as the top-level host. Connect the Claude Code CLI to DSH via the Agent Communication Protocol (ACP) or a community bridge (like `dsh-claude-live`).
* **Strict Billing Compliance:** The setup must execute using my local `claude login` session token to consume my $200 Max Plan quota. Do not use raw Anthropic API keys that bypass this plan.
* **Context Optimization:** Configure DSH's Code Mode SDK to batch multi-step file operations into single passes, preventing Claude Code from burning context on native 15-turn tool loops.
* **Mem0 (or alternative) (Long-Term Memory) Integration:** Install and configure the Memory framework. You must wire it into DSH as a Cordis plugin so the agent can store user directives, architecture decisions, and project conventions in a vector store and retrieve them dynamically without bloating the system prompt.
**Expected Engagement**
* **Initial Setup (4–5 hours):** Establish the baseline DSH-to-Claude connection, verify token routing, deploy the Mem0 storage backend, and test the memory retrieval hooks.
* **Ongoing Support (1–2 hours/week):** Maintain plugins, resolve ACP sub-process lag, and refine the Mem0 extraction rules to keep context retrieval highly relevant.
**Ideal Candidate**
You must have proven experience with the Cordis microkernel, terminal I/O bridging, and vector memory systems (like Mem0 or similar RAG implementations). Please start your proposal with "DSH-MEM0" so I know you read the requirements.
---
### DeepSeek Harness (DSH) & Claude Code Integration Expert (with Memory plugin Setup)
**Project Overview**
I need a developer to configure a meta-agent workspace integrating DeepSeek Harness (DSH), Claude Code, and Mem0. The goal is to use DSH as the workspace orchestrator to handle prompt caching and file execution, route the actual reasoning through Claude Code, and use Mem0 to retain long-term project memory across sessions.
**Core Technical Requirements**
* **Architecture & Routing:** Deploy DeepSeek Harness (`npx @deepseek-ai/dsh web`) as the top-level host. Connect the Claude Code CLI to DSH via the Agent Communication Protocol (ACP) or a community bridge (like `dsh-claude-live`).
* **Strict Billing Compliance:** The setup must execute using my local `claude login` session token to consume my $200 Max Plan quota. Do not use raw Anthropic API keys that bypass this plan.
* **Context Optimization:** Configure DSH's Code Mode SDK to batch multi-step file operations into single passes, preventing Claude Code from burning context on native 15-turn tool loops.
* **Mem0 (or alternative) (Long-Term Memory) Integration:** Install and configure the Memory framework. You must wire it into DSH as a Cordis plugin so the agent can store user directives, architecture decisions, and project conventions in a vector store and retrieve them dynamically without bloating the system prompt.
**Expected Engagement**
* **Initial Setup (4–5 hours):** Establish the baseline DSH-to-Claude connection, verify token routing, deploy the Mem0 storage backend, and test the memory retrieval hooks.
* **Ongoing Support (1–2 hours/week):** Maintain plugins, resolve ACP sub-process lag, and refine the Mem0 extraction rules to keep context retrieval highly relevant.
**Ideal Candidate**
You must have proven experience with the Cordis microkernel, terminal I/O bridging, and vector memory systems (like Mem0 or similar RAG implementations). Please start your proposal with "DSH-MEM0" so I know you read the requirements.
---
Apply on Freelancer →
Project sourced from Freelancer.com. Applications happen directly on the original platform — we never collect your data.