API Projects

Projects API

Projects are discovered by reading ~/.claude/projects/. Each subdirectory corresponds to a working directory where Claude Code was used.

Queries

projects

List all discovered projects.

{
  projects {
    encodedName
    path
    sessionCount
    transcriptCount
    toolUsage { name count }
    skillsUsed { name count }
  }
}

Field	Description
encodedName	Raw directory name (e.g. -Users-alice-go-src-github-com-alice-myrepo)
path	Decoded filesystem path (e.g. /path/to/your/project)
sessionCount	Sessions recorded in the DB with this CWD
transcriptCount	.jsonl transcript files found on disk
toolUsage	Top tools used across all sessions for this project
skillsUsed	Skills invoked across all sessions for this project

project

Fetch a single project by filesystem path.

{
  project(path: "/path/to/your/project") {
    encodedName
    path
    sessionCount
    sessions(limit: 10, offset: 0) {
      id
      hookCount
      durationSeconds
      lastSeenAt
    }
  }
}

Path decoding

Claude Code encodes project paths by replacing / and . with - in the directory name. This encoding is lossy — a path like /Users/alice/my-project and /Users/alice/my/project would encode the same way. The server uses a depth-first filesystem search to reconstruct the most likely path, but short or ambiguous names may decode incorrectly.

Project intelligence

Projects aggregate session intelligence by working directory. This is a good way to answer “what tends to happen in this repo?”

{
  project(path: "/path/to/your/project") {
    path
    sessionCount
    transcriptCount
    skillsUsed { name count }
    toolUsage { name count }
    sessions(limit: 10) {
      id
      status
      lastSeenAt
      model
      editedFiles
      prompts
    }
  }
}

userSkills

List skills discovered in ~/.claude/skills/:

{
  userSkills {
    name
    description
    dirName
  }
}

Compare discovered skills with skillsUsed on sessions or projects to see which skills are actually being invoked.