Show HN: A LinkedIn / X CLI for your agents

CyrusNewDay launched a CLI tool called 'social' that lets AI agents interact with LinkedIn and X accounts via JSON commands, with caching and local mirroring to reduce costs. The tool is designed for agent-first usage, supporting deterministic exit codes and structured errors.

Accounts — social account connect linkedin · social account reconnect linkedin <account · social account disconnect linkedin <account · social account billing · social account. Connect, re-auth, inspect billing, and list the accounts on this network. Accounts — social account connect x · social account reconnect x <account · social account disconnect x <account · social account billing · social account. Connect, re-auth, inspect billing, and list the accounts on this network. Composable with the tools you already use. Every command speaks JSON. Pipe through jq , fzf , your agent's planner — anything that reads stdin. Shared command patterns across both networks. --limit bounds pages, LinkedIn list reads use --offset , and offset responses can expose .meta.totalCount . Cursor-based reads use --cursor and .meta.cursor ; cacheable reads also take -H "Cache-Control: no-cache" for fresh upstream reads. Repeated exploration can get cheaper. Allowlisted GET reads can hit cache when your agent asks the same question again, so those repeat hits cost $0 . Designed for agents first. Deterministic exit codes, structured errors, and a SKILL.md you drop straight into Claude, Codex, or any other agent.Agents tend to ask the same read questions while exploring. social has two read layers: live proxy reads, where selected repeats can hit cache, and the local mirror, where synced data can be queried for free with SQL. sync pulls your own data into the local mirror and can spend usage. sql queries that mirror read-only, returns the standard envelope, and omits meta.cost . 15 minutes . Change it with social account config cache ttl {total in seconds} . -H "Cache-Control: no-cache" on cacheable read commands when freshness matters. It skips the cached read and refreshes the stored response after a successful upstream call. no-store bypasses cache read and write; max-age=<seconds overrides the TTL for one request. ETag and Last-Modified . The request cache surface is still Cache-Control ; use no-cache for a fresh upstream read. meta.cache.tables , including last sync time and age. mode 0600 . Stored in your OS keyring when one is available; this file is the fallback. social account login : read,write default or read . Carried in the token and revocable with social account logout — there is no ambient write access. Cache-Control: no-cache , no-store , or max-age=<seconds when the command lists --header . read,write . Clear Write in the prompt for a session that can't post, invite, or follow. $0 . Audit it any time with social account billing , social account usage , and social account logs — nothing happens on your accounts you can't see. LinkedIn Account Safety /security/linkedin . Built by @CyrusNewDay https://x.com/cyrusnewday because it's a tool he needed.