DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5

How to tune Serena's tools and memory per project with Serena

Use Serena's layered YAML configuration and memory system to fit each repository and client.

Short answer: Install Serena with uv tool install -p 3.13 serena-agent; then Adjust global configuration, then add per-project configuration with local overrides; use execution context-specific configuration for particular clients and modes for composable fragments. All steps.

Who it is forA team lead rolling Serena out across several repos and clients who wants different tools active in each.

The problemOne global setup rarely fits every project, and duplicated tools confuse agents that already have file and shell tools.

The outcomePer-project Serena settings with the right tools active, and memories shared across sessions alongside AGENTS.md files.

What you need first

  • uv, the only required prerequisite
  • An MCP client such as Claude Code, Codex, OpenCode, Gemini-CLI, Cursor, VS Code or a JetBrains assistant
  • An LLM to drive the tools, since Serena itself does no reasoning
  • For some languages, extra language server dependencies listed on the Language Support page
  • Optional: the paid Serena JetBrains Plugin (free trial) for move, inline and debugging

Steps

  1. Install Serena with uv tool install -p 3.13 serena-agent
  2. Adjust global configuration, then add per-project configuration with local overrides
  3. Use execution context-specific configuration for particular clients and modes for composable fragments
  4. Leave basic tools such as search_for_pattern and execute_shell_command disabled inside harnesses that already provide them
  5. Turn on the memory system, or disable it if you use another, and combine it with your AGENTS.md files

Prompt for your agent

Configure oraios/serena for this repository: per-project configuration with only the retrieval and symbolic editing tools active for <client>, the memory system on, and a short note in AGENTS.md on how agents should use Serena memories.

Caveat. The README describes the configuration levels but the YAML keys are in the docs, so check the configuration page for exact names.

Use this in your AI agent

Ask your agent

Using the Deduped MCP, get the setup steps for oraios/serena and "Tune Serena's tools and memory per project"

Add Deduped to Claude Code once

claude mcp add --transport http deduped https://dedupedai.com/mcp

Cursor, ChatGPT, others

Share: X LinkedIn Hacker News Email Copy link

Third-party repos we don't own. AI-written use cases can be wrong. Disclaimer