The same actions,
straight from the shell.
Pipe-friendly commands for humans and agents running in shell mode. One binary, one token, one project graph — same backend as MCP, same IDs, same audit trail.
One binary. Same auth token.
Install the CLI, sign in with a personal access token, and pin the active space. The same PAT works across MCP, REST, and the CLI — revoke once and every surface stops talking.
Three lines to a working CLI
Same script for fresh laptops and rebuilt containers.
brew install hydrant/tap/hydrant
hydrant auth login --pat hyd_sk_…
hydrant use <space>Every Hydrant primitive, mapped to a verb.
The CLI mirrors the MCP tool catalog one-for-one — issues, decisions, planning, notes, files, and context assembly. Subcommands stay grouped so muscle memory transfers between agents and shells.
Issues
hydrant issue createhydrant issue listhydrant issue updatehydrant deps setDecisions
hydrant decision addhydrant decision listhydrant decision linkPlanning
hydrant cycle currenthydrant cycle listhydrant milestone createNotes
hydrant note createhydrant note listhydrant folder createFiles
hydrant file savehydrant file listhydrant file urlContext
hydrant searchhydrant bundlehydrant preflightFrom idea to preflight in five lines.
Spin up an issue, capture the constraint as a decision, lock the dependency, bundle the context, and run preflight — all without leaving the shell. Same identifiers your agent will see when it picks the work up.
Mint a token, bind it to the shell, revoke it when it's done.
Personal access tokens are the only credential the CLI needs. Generate one in settings, paste it into hydrant auth login, and rotate or revoke without touching the rest of your stack.