Practical overview
When it fits
Use Dolphin when an LLM application needs persistent memory beyond a single prompt, or when Claude Code sessions should reuse project context. Choose the SDK for application-managed, user-specific memory and the plugin or MCP server for agent access.
Who it is for
- Developers adding persistent, user-scoped memory to an LLM application.
- Developers who want relevant stored context to be retrieved for prompts.
- Claude Code users who want project decisions and conventions to carry across sessions.
Documented uses
- Store application memories under a user-specific namespace and retrieve relevant context.
- Recall semantically related memories and graph relationships for an LLM prompt.
- Share project-scoped memory between Claude Code sessions.
- Expose memory operations through the documented MCP server.
Requirements & limits
The project is open source under the MIT license. The README specifies Python 3.10 or later and notes that the described 0.2.0 source version is not on PyPI; its `pip install dolphin-memory` command is described as installing 0.1.0, which requires Supabase.
- Storage choices differ by use: local SQLite is documented for local SDK use and shared agent sessions; Supabase is documented for server-side SDK applications.
- Memory storage and search are described as working out of the box, while knowledge-graph construction requires an LLM.
- The SDK retrieves context for the application to inject into its system prompt; Dolphin is a memory layer rather than the LLM application itself.
- The documented plugin workflow is for Claude Code; MCP is also offered for compatible agent clients.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Install Dolphin from source for local SDK use
Set up the documented source version of the Python SDK, which includes local SQLite storage.
Before you start
- Python 3.10 or later
- Access to the Dolphin GitHub repository
Steps
Install the SDK from the repository’s main branch using the README’s pip source-install command.
Initialize
DolphinMemory()in a Python application; the README says local SQLite is the default.Use the SDK’s
add,searchorget_contextoperations for application memory.
Expected outcome
A Python application can use Dolphin’s documented memory operations with local SQLite storage.
Guide limitations
- The README says the 0.2.0 version with local SQLite is not on PyPI;
pip install dolphin-memoryinstalls 0.1.0, which requires Supabase. - Building the knowledge graph requires an LLM; Ollama is the stated default local option.
Add shared memory to Claude Code
Set up the documented Dolphin plugin so Claude Code sessions can use shared project memory.
Before you start
- Python 3.10 or later
- Claude Code
- Access to the Dolphin GitHub repository
Steps
Install
dolphin-memoryusing the README’s documented pip command.Add the Dolphin marketplace with
/plugin marketplace add DewashishCodes/dolphin.Install the plugin with
/plugin install dolphin-memory@dolphin.
Expected outcome
The documented plugin adds project memory to session context, recalls related memories for prompts and stores selected decisions, conventions and gotchas.
Guide limitations
- The described shared memory is scoped to a project’s Git remote.
- The README names Claude Code for this plugin workflow; it does not document the same plugin setup for other agent clients.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 14 October 2026
- Original website title
- DewashishCodes/dolphin · Official homepage title · retained in its source language