MoonshotAI / MoonshotAI/kimi-code
Feature request: General memory provider mechanism for automatic cross-session memory integration
Nobody has claimed this yet.
- Dominant language
- TypeScript
- Stars
- 7.5k
- Forks
- 1.2k
- Avg merge
- 11h 53m
- Merged PRs (30d)
- 350
Description
Kimi Code CLI already supports MCP servers and lifecycle hooks, but there is no built-in, backend-agnostic memory provider mechanism comparable to Hermes Agent's memory provider plugin or OpenClaw's memory integration.
Use case
Users who run memory systems such as OpenViking would like important facts, preferences, decisions, and project context from a session to be automatically extracted, stored, and recalled in future sessions — without having to call mcp__openviking__remember / mcp__openviking__recall manually every turn.
Proposed solution
Add a general memory provider interface/plugin point that:
- Allows declaring one or more memory providers (MCP-based or native).
- Automatically extracts memories at configurable trigger points (e.g., session end, context compaction, explicit command).
- Injects relevant memories into context at configurable points. Injection should be available at least at session start and on a per-turn basis, so the agent can recall context exactly when it is needed.
- Is backend-agnostic, so any memory provider can implement it.
Prior art
- Hermes Agent: dedicated memory provider plugin for OpenViking.
- OpenClaw: similar automatic memory extraction/recall.
Current workaround
Manual MCP tool calls (mcp__openviking__remember, mcp__openviking__recall) or a custom plugin hook that talks to the memory service externally. This is tedious and cannot provide the same deep, general integration as a first-class provider mechanism.
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Research direction
The issue does not identify files, tests, or entry points; start by locating the CLI's MCP, lifecycle-hook, session, and context-compaction integrations. Done means defining a backend-agnostic provider mechanism that supports configurable extraction and memory injection at session start and per turn.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- typescript
- Domain
- ai, cli
- Issue type
- Feature
- Difficulty
- 5/5
- Estimated time
- Over a week
- Activity status
- Quiet
- Clarity
- Needs clarification
- Newbie friendliness
- 25/100