Best prompt manager for Claude Code (2026)

We build PromptLikeEngineer, one of the five options below. Competitor prices are as listed on each product's own site, last checked on October 11, 2026. Pricing moves, so check again before you buy.

Three levels, from weakest to strongest:

1. Copy-paste compatible  any web tool; you move the text by hand
2. CLI-reachable          a command pulls the prompt; Claude doesn't know the library exists
3. Agent-reachable        Claude Code calls a tool (MCP) or loads a skill that
                          fetches the prompt itself, mid-conversation

Keep this ladder in mind while reading feature lists. It is the axis that decides how much copying you do every day.

Versioning, visual diffs, side-by-side tests across models and a bring-your-own-key setup, in a clean web editor. Genuinely good if you write and refine prompts in a browser. Its site lists no CLI or MCP server, so getting a prompt into Claude Code means copying it out by hand. It is $9/mo billed annually, with a 14-day trial and no free plan.

Pitches itself as a shared workspace for prompts and backs it with a low paid tier. The free plan keeps prompts public; private prompts start on Pro. There is a REST API (limited on Free, full on Pro), so you can wire it into a script, but its pricing page lists no CLI or MCP server. A reasonable pick if your prompts and your Claude Code sessions are separate parts of your day. We wrote a head-to-head comparison.

Open source and strong at tracing and observability for an application that calls an LLM in production. Prompt management exists inside it, but it is secondary to that job. If your question is "what did my deployed app do, and what did it cost," Langfuse is the right tool and this comparison isn't about you. If your question is "where's the prompt I wrote yesterday," it is more tool than you need.

The honest baseline: a folder of .md files in git is free, greppable, diffable and yours. Claude Code reads files directly, so this is arguably the simplest way to get a prompt into a session. The cost shows up later: no search across dozens of files, no {{variable}} filling, no sync beyond your own git discipline, and nothing that lets Claude reach for a prompt by name without you naming the file. The full trade-off is in PromptLikeEngineer vs dotfiles/Gist.

The differentiator for this query is narrow. PromptLikeEngineer is the one option here that Claude Code can query by itself, mid-session: the MCP server exposes list_prompts, search_prompts, get_prompt and fill_variables, so Claude can check your library before writing a long prompt from scratch. The same prompt is one ple get away in the terminal and versioned in one place.

npm i -g promptlike && ple login
claude mcp add promptlike -- npx -y promptlike-mcp

Where it is weaker: billed yearly, Pro works out to $15/mo against $9/mo for PingPrompt and PromptHub (or $19 against PromptHub's $12 billed monthly). If budget decides it and you don't need the agent surface, those two are the cheaper honest answer. If self-hosting matters more than convenience, look at the open-source options in our broader comparison.

Browser only, lowest price             → PingPrompt or PromptHub
Production tracing for an app          → Langfuse
Zero tools, full control               → a dotfiles folder / Gist
Claude Code fetches the prompt itself,
plus CLI and web                       → PromptLikeEngineer

Want to see the agent-reachable level for yourself? The MCP setup takes two commands, and the Free plan covers 25 prompts with no card.