Everything an AI agent needs to store data on Filecoin Onchain Cloud and read it back, in one place
Paste one into Claude Code, Cursor, Codex or any coding agent. Each was run through an agent before it was put here; the note under each says how far.
Publish the folder ./my-folder to Filecoin Onchain Cloud and give me a link I can share. This is mainnet and uses real funds. 1. Install the official publish skill into this project (needs Node 24 or newer): npx skills add filecoin-project/filecoin-skills --skill publish 2. Read the installed skill's SKILL.md and follow it. If it needs me to log in or add funds, stop and tell me exactly what to do. 3. I am asking for a cost preview before any real upload: run the dry run with the same options the real publish will use. Tell me the deposit needed now, the one-time fees and the monthly storage rate, and list the files that will become public. 4. Do not publish or move any funds until I say yes. 5. After I say yes: publish, wait until the link is confirmed working, and give me the link.
Uses the official publish skill, which writes to mainnet and needs a funded wallet. Tested as far as the cost preview; the agent then waits for your yes.
Read a file back from Filecoin Onchain Cloud and check it. The file's IPFS CID is bafybeiapnvjfxw2w7556sb7ixrhhy26nk6p36tmuh27hkysaqiw4kpr7my. 1. Ask the public index which storage providers hold it: GET https://cid.contact/cid/<CID>. Provider addresses are under MultihashResults[].ProviderResults[].Provider.Addrs and look like /dns/HOST/tcp/443/https, which means https://HOST. 2. Download it from one of those providers as a CAR file: GET https://HOST/ipfs/<CID> with the header "Accept: application/vnd.ipld.car". If one provider fails, try the next. 3. Check that every block in the CAR hashes to its own CID and that the CAR's root is the CID I gave you. Then unpack the file from the CAR (it is UnixFS; a tool such as ipfs-car or ipfs-unixfs-exporter does this). 4. Tell me the unpacked file's type, its size in bytes and its SHA-256.
Needs no wallet and costs nothing. Swap in any CID; this one is a photograph from the archive on this site. Tested end to end.
Made by the teams named on each. focify.me only points at them.
Demos where an agent did the work, or where an agent is the intended user
The official documentation publishes a plain-text index for language models at docs.filecoin.cloud/llms.txt. This site has its own, describing what is here: focify.me/llms.txt.