Bear. CTXPM

Bear. CTXPM (Context Package Manager) is an open-source protocol created by Bear. Best, an indie developer, for organizing AI resources in software projects. It treats the skills, rules, prompts, memories, and MCP configurations that AI coding agents rely on as first-class project assets with clear ownership, versioning, and lifecycle rules. The project is actively maintained on GitHub and reached v0.1.16 as of September 2026.

What It Is

Bear. CTXPM is a protocol-first system for managing AI resources in development projects. As AI coding agents become part of everyday engineering workflows, projects accumulate two kinds of AI resources: external shared resources (team-wide or third-party) and project-local resources tightly coupled to the codebase. Without a convention, these files scatter across the project with no clear ownership or update boundary. Bear. CTXPM solves this by introducing unified dependency / package semantics, a standard directory structure under .ctxpm/, a ctxpm.yaml manifest, and a companion ctxpm CLI for repeatable operations.

Two Ownership Concepts

The protocol exposes exactly two ownership terms to users:

• dependency: an external AI resource the project uses but does not maintain. Its source, path, and version are recorded in ctxpm.yaml; the content typically lives outside version control.
• package: an AI resource maintained by the current project, committed to version control, and reviewed alongside source code.

Each resource also has a type—skill, rule, spec, prompt, memory, or mcp—which describes its content shape independently of ownership. This combination (e.g., dependency + skill or package + rule) makes it immediately clear who maintains a resource and where to make a fix.

Protocol-First Architecture

Bear. CTXPM is designed so a project does not depend on a binary before an AI can understand its working environment. The protocol lives in three places:

• ctxpm.yaml — the project manifest declaring all resources, their sources, types, and agent profiles
• .ctxpm/ — the directory structure separating dependencies/ (gitignored) from packages/ (committed)
• Agent entrypoint documents — a canonical .ctxpm/AGENTS.md file, with root-level symlinks (AGENTS.md, CLAUDE.md, GEMINI.md, etc.) pointing to it so each agent reads consistent instructions

The CLI handles fixed, repeatable steps: ctxpm init, ctxpm add, ctxpm list, ctxpm validate, ctxpm detect, ctxpm check-updates, and ctxpm update. Discovery and reporting are always separated from applying changes, so AI can surface a problem while the user decides whether to act.

Standard AI Workflow

When an AI agent enters a Bear. CTXPM project, the recommended workflow is:

1. Read the root entrypoint (AGENTS.md or equivalent symlink) and ctxpm.yaml
2. Scan .ctxpm/packages/ first, then .ctxpm/dependencies/
3. Identify resources from explicit declarations and directory structure
4. Load memory resources on demand when a task depends on project history
5. Maintain the managed ctxpm block in .ctxpm/AGENTS.md and keep symlinks current
6. Report safely when an operation cannot be completed, and provide migration suggestions

For version tracking, Bear. CTXPM records the last commit that changed a resource path rather than the repository's latest HEAD, preventing misleading update signals when unrelated directories change.

Update: v0.1.16

The latest release is v0.1.16, published on September 11, 2026. The repository was created in July 2026 and has seen active iteration through the v0.1.x series. The project is written primarily in Go and is hosted on GitHub under the gBearBest account. The GitHub README notes that v0.1 explicitly does not require a self-hosted registry, a lockfile mechanism, or dedicated metadata files per resource—keeping the protocol lightweight and agent-agnostic.

Who It Is For

Bear. CTXPM is aimed at developers and teams who use AI coding agents on projects that need long-term maintenance. It is most useful when multiple agents need to read the same project rules, when teams want to reuse shared skills without copying them into every repository, when project-owned prompts and specs need code review, or when old AI files are scattered and need inventorying before migration. One-off AI demos may not need it, but any project with ongoing AI resource accumulation benefits from the explicit ownership and lifecycle model it provides.