Context Keeper
Persist project decisions, pipelines, and constraints so Claude picks up the same context every session instead of re-deriving your stack from scratch.
Overview
Context Keeper is a MCP server for the Build phase that stores project decisions, pipelines, and constraints so Claude retains memory across conversations.
What is this MCP server?
- Project-scoped memory for decisions, pipelines, and constraints across Claude conversations
- context-keeper-mcp on PyPI with stdio MCP transport (v0.1.1)
- CONTEXT_KEEPER_PROJECT env points memory at an absolute project directory
- Reduces repeated re-explaining of stack choices, naming, and workflow gates
- Designed for long-horizon solo builds where context windows reset frequently
- PyPI identifier context-keeper-mcp version 0.1.1
- 1 primary configuration env var CONTEXT_KEEPER_PROJECT
- stdio transport; defaults project path to current working directory if unset
What problem does it solve?
Every new agent session forces solo builders to re-upload context, re-state decisions, and watch the model violate constraints that were settled weeks ago.
Who is it for?
One-person teams on multi-week builds who live in Claude Desktop and need persistent decision and constraint recall per repository.
Skip if: Organizations with mandatory Notion or ADR workflows that already feed every agent session automatically.
What do I get? / Deliverables
After installing Context Keeper MCP, Claude reads and updates project memory tied to your repo path so implementation sessions stay aligned with prior decisions and pipelines.
- Durable store of project decisions readable by the agent via MCP
- Documented pipelines and constraints referenced in later sessions
- Reduced re-onboarding time at the start of each Claude conversation
Recommended MCP Servers
Journey fit
Context Keeper is shelved under Build PM because capturing decisions and constraints is where solo builders most often lose thread mid-implementation. PM subphase fits durable project memory—architecture choices, pipeline rules, and non-negotiable constraints—that agents must honor while coding.
How it compares
Project memory MCP server, not a task runner, identity layer, or marketplace skill bundle.
Common Questions / FAQ
Who is io.github.jarmstrong158/context-keeper for?
Solo builders and small teams using Claude who need cross-session memory for decisions, pipelines, and project constraints in a specific codebase.
When should I use Context Keeper MCP?
Use it as soon as a project has non-obvious decisions and recurring pipelines—typically during Build and whenever you iterate in Operate—so agents do not reset context each chat.
How do I add Context Keeper to my agent?
Install context-keeper-mcp from PyPI, register stdio MCP in your host, and set CONTEXT_KEEPER_PROJECT to the absolute path of the project directory you want remembered.