Eve Horizon CLI

Published CLI (npx) for interacting with the Eve Horizon API. Dev/ops tooling lives in ./bin/eh.

npx @eve/cli --help

Philosophy

The CLI is the primary interface for Eve Horizon, designed for humans and AI agents. The REST API is the substrate: every operation is exposed via HTTP, and the CLI is a thin wrapper that handles argument parsing and output formatting. It does not bypass the API or contain business logic.

See API Philosophy and OpenAPI.

Profiles (defaults + credentials)

Profiles store API URL, default org/project IDs, default harness, and Supabase auth config.

# Create or update a profile
eve profile set local \
  --api-url http://localhost:4801 \
  --org org_defaulttestorg \
  --project proj_xxx

# Set default harness (with optional variant)
eve profile set --harness mclaude
eve profile set --harness mclaude:fast   # harness with variant

# Set active profile
eve profile use local

# Show active profile
eve profile show

Profile harness defaults are used when scheduling jobs. The --harness flag accepts harness or harness:variant format everywhere. Priority: --harness flag > job hints > profile default > system default.

Supabase config (cloud auth)

Store Supabase settings in a profile (recommended for cloud stacks):

eve profile set prod \
  --api-url https://api.eve-horizon.example.com \
  --supabase-url https://your-project.supabase.co \
  --supabase-anon-key <anon-key>

Auth

Auth is required for cloud stacks. The default flow uses GitHub SSH keys; Supabase remains an optional adapter for legacy deployments.

The CLI will refresh Supabase access tokens automatically if a refresh token is available.

# SSH login (default)
eve auth login --email [email protected] --ssh-key ~/.ssh/id_ed25519

# Supabase login (optional)
eve auth login --email [email protected] --password '...' \
  --supabase-url https://your-project.supabase.co \
  --supabase-anon-key <anon-key>

# Status / whoami
eve auth status
eve auth whoami

# Logout
eve auth logout

Commands

Organizations

eve org ensure "My Org"
eve org list
eve org get org_xxx
eve org update org_xxx --name "New Name"

Projects

eve project ensure --name my-project --slug MyProj --repo-url https://github.com/org/repo
eve project list
eve project get proj_xxx

Jobs

Jobs follow a phase-based lifecycle: idea → backlog → ready → active → review → done

Jobs created without a --phase flag default to ready, making them immediately schedulable.

# Create a job (defaults to ready phase)
eve job create --description "Fix the login bug in auth.ts"
eve job create --description "Add dark mode" --priority 1 --harness mclaude

# Create a job with git controls
eve job create \
  --description "Fix checkout" \
  --git-ref main \
  --git-branch job/fix-checkout \
  --git-create-branch if_missing \
  --git-commit auto \
  --git-push on_success

# List and filter jobs
eve job list --phase ready
eve job ready                    # Schedulable jobs (ready, not blocked)
eve job blocked                  # Jobs waiting on dependencies

# View job details
eve job show MyProj-abc123
eve job tree MyProj-abc123       # Hierarchical view

# Update jobs
eve job update MyProj-abc123 --phase active --assignee agent-1
eve job close MyProj-abc123 --reason "Work completed"
eve job cancel MyProj-abc123 --reason "No longer needed"

# Dependencies
eve job dep add MyProj-abc123 MyProj-def456   # abc123 depends on def456
eve job dep list MyProj-abc123

# Claim/release workflow (typically used by scheduler/agents)
eve job claim MyProj-abc123 --harness mclaude
eve job release MyProj-abc123 --reason "Need more info"
eve job attempts MyProj-abc123

# Job execution and monitoring
eve job logs MyProj-abc123
eve job result MyProj-abc123 --format full    # Get job results (text|json|full)
eve job result MyProj-abc123 --attempt 2      # Get results from specific attempt
eve job wait MyProj-abc123 --timeout 300      # Wait for job completion
eve job wait MyProj-abc123 --quiet --json     # Wait quietly, JSON output
eve job follow MyProj-abc123                  # Stream logs in real-time via SSE
eve job follow MyProj-abc123 --raw            # Stream raw JSON lines
eve job follow MyProj-abc123 --no-result      # Stream logs without final result

# Review workflow
eve job submit MyProj-abc123 --summary "Implemented fix"
eve job approve MyProj-abc123 --comment "LGTM"
eve job reject MyProj-abc123 --reason "Missing tests"

Job Results

Fetch and display completed job results:

# Show full job result (default format)
eve job result MyProj-abc123

# Get results in different formats
eve job result MyProj-abc123 --format text    # Plain text output only
eve job result MyProj-abc123 --format json    # Full JSON structure
eve job result MyProj-abc123 --format full    # Formatted with metadata (default)

# Get results from a specific attempt
eve job result MyProj-abc123 --attempt 2

Waiting for Job Completion

Block until a job completes, with optional timeout:

# Wait for job to complete (default timeout: 300s)
eve job wait MyProj-abc123

# Custom timeout (max 300s enforced by API)
eve job wait MyProj-abc123 --timeout 120

# Quiet mode (no progress output)
eve job wait MyProj-abc123 --quiet

# JSON output format
eve job wait MyProj-abc123 --json

Exit codes:

• 0: Job completed successfully
• 1: Job failed
• 124: Timeout reached
• 125: Job was cancelled

Real-time Log Streaming

Stream job logs in real-time using Server-Sent Events (SSE):

# Stream logs with timestamps and formatted output
eve job follow MyProj-abc123

# Stream raw JSON lines (for parsing/filtering)
eve job follow MyProj-abc123 --raw

# Stream logs without printing final result
eve job follow MyProj-abc123 --no-result

The follow command connects to the job's SSE endpoint and displays logs as they are generated. It shows:

• Timestamps for each log entry
• Tool names and actions

Secrets

Secrets can be stored at user/org/project scope. Values are never returned in plaintext; show returns a masked value.

# Project secret
eve secrets set GITHUB_TOKEN ghp_xxx --project proj_xxx --type github_token
eve secrets list --project proj_xxx
eve secrets show GITHUB_TOKEN --project proj_xxx
eve secrets delete GITHUB_TOKEN --project proj_xxx

# Import host secrets from .env
eve secrets import --project proj_xxx --file .env

• Formatted, human-readable output

Use --raw for programmatic consumption or when piping to other tools.

Scheduling Hints

Jobs can include hints for the scheduler:

eve job create --description "Heavy computation" \
  --harness mclaude:fast \
  --worker-type gpu \
  --permission auto_edit \
  --timeout 7200

Creating Sub-Jobs (for agents)

Agents can create and optionally claim sub-jobs inline:

# Create sub-job under parent
eve job create --parent MyProj-abc123 --description "Implement feature X"

# Create and immediately claim (for inline execution)
eve job create --parent $EVE_JOB_ID --description "Sub-task" --claim

Environment variables for agent context:

• EVE_PROJECT_ID - Current project
• EVE_JOB_ID - Current job being executed
• EVE_ATTEMPT_ID - Current attempt UUID
• EVE_AGENT_ID - Agent identifier

Harnesses

eve harness list              # List available harnesses
eve harness get mclaude       # Show harness details and auth status
eve harness list --capabilities  # Include model/reasoning capability hints

Agents (Orchestration Config)

Provide policy + harness context for orchestrating agents (profiles, councils, availability):

eve agents config --json
eve agents config --path /path/to/repo --no-harnesses

Recommended default policy profile name: primary-orchestrator.

Pagination

List endpoints accept --limit and --offset. Default limit is 10, newest first.

eve job list --limit 10 --offset 20

Dev CLI

Local dev/ops tooling:

./bin/eh --help
./bin/eh dev start
./bin/eh k8s start     # Default runtime
./bin/eh docker start  # Quick dev loop