FixMyPrompt started as a prompt checker and has grown into something closer to a memory for your AI. You can still paste a prompt and get a score out of 100 with a fixed version, but the bigger idea now is that when your coding agent finally works something out, it saves how, so the next session follows those steps instead of spending another hour on the same problem.
Disclosure: FixMyPrompt is our own product, and you will see it recommended across this site. It is reviewed here with the same honest read we give every tool, cons included.
Memories: the part that changed
A FixMyPrompt memory is more than a saved prompt. When your AI gets something right, you say "save how you fixed that to memory," and it stores the steps, the commands, the check that proves it worked, and the mistakes to avoid. A secret check runs first, so keys and passwords are caught before anything is stored. Memories are filed into a project named after the folder you are working in, so one repo's steps do not leak into another.
Using one later is plain language. In Claude Code you ask "use fmp to find how we deploy staging," and your AI reads back the memory that fits and follows it. Give a memory a short name and it also runs as a slash command, asking you to fill in any blanks like a branch name or environment each time. After a run, your AI reports whether the steps worked, and a memory that failed gets marked so somebody fixes it. Every change is a new version with a mark on the one in use, and on Pro you can undo back to an earlier one.
The honest limit is that your AI does not go looking for memories on its own yet. You still ask, or run the command. On Pro, Claude Code and Codex can save lessons as drafts while you work and import lessons from past sessions on your machine, and every draft waits for you to keep, edit, or delete it. Only the lessons are sent, never your code.
Prompt Fix and Agent Doctor
The original prompt check is still free for 3 checks a day with no account. It reads a question, a system prompt, an agent task, and a workflow differently, because each one needs different things to work. The check I find most useful is the one that tells you a prompt cannot run at all, no matter how it is worded, instead of handing back a polished version of something your tool can never do. Tighten trims a long saved prompt and shows the real token count before and after.
Agent Doctor applies the same idea to the CLAUDE.md or AGENTS.md file your coding agent reads at the start of every session. Paste one in and it shows what the file costs in tokens each session, the rules it is missing, and blocks an installer pasted in. On Pro, the full doctor checks several agent files together for rules that clash and hands back a fixed file with every change listed, and you can run it inside Claude Code or Codex without sending your source code. A GitHub check can also comment a score on every pull request that changes an agent file.
Team Brain on Biz
Biz is for a company that wants everyone's AI to reuse what one person already worked out. A memory one person saves can be found by a teammate's AI in their own session, and changes to shared memories wait for an admin to approve them. When someone leaves, their seat frees up and what they saved stays with the company. Biz also adds a company rules page and optional role pages loaded into every session, with locked lines that cannot be overridden, which doubles as onboarding for a new hire's AI. Enterprise, from 200 seats, adds project leads and a single review page across projects.
Pricing and what you get
The full breakdown is on the pricing page. Free costs nothing and connects one AI tool with 1 project, 25 memories, 25 session notes, 300 memory reads a month, and 5 GitHub checks a month. Free memories are kept 60 days after you last use them, and prompt checks on Free are paid with credits.
Pro is $19/month, with a 7-day free trial that needs a card. It connects every AI tool you use to the same memory, raises you to 5 app projects plus folder projects and 5,000 memories kept 6 months after last use, and includes 20,000 credits a month for prompt checks and Tighten. Biz is $25 a seat per month with a 2-seat minimum, adds the company sharing above, and pools 20,000 credits per seat across the company. Enterprise is custom pricing from 200 seats.
Credit packs run from $2.99 to $21.99, work on any plan, and never expire. A $2.99 pack covers about 44 basic checks, which run on Claude Haiku 4.5, with deeper checks on Sonnet 5.5 and Opus 5.5 costing more per check. Any plan, Free included, can plug in your own OpenRouter key so AI runs bill to your key and cost no credits.
Where it falls short
The memory side is built around coding agents. It connects to Claude Code over MCP with one command, and the session tools are for Claude Code and Codex, so if you only ever use a chat app in the browser, most of the value here does not reach you. You can import lessons from chat apps by pasting or uploading an export, but that is a one-way trip rather than a live connection. A few features on the roadmap, like asking the team memory from Slack and single sign-on, are labeled as coming and are not available yet. And because the AI does not look memories up by itself, the habit of asking for them is on you.
Who it fits
FixMyPrompt fits developers and technical teams who use Claude Code or Codex every day and keep re-explaining the same setup, deploy, or debugging steps to a fresh session. The Free plan is enough to see whether saving and reusing memories changes how your week goes. Biz is worth a look for a team where one person always ends up being the one who knows how something works. If you only want a one-off prompt score, skip the account and use the free checker, and if you never touch a coding agent, a general note-taking tool will serve you better.