Cross-Repo Search

By default, tribal recall searches memories scoped to the current project. With the --all flag, you can search across every project (assistant) in your Backboard account.

When to use cross-repo search

• You know knowledge exists but aren’t sure which project it’s in
• You’re working in a monorepo-adjacent setup with multiple repos that share patterns
• You want to find how another project solved a similar problem
• An agent needs context about a term or concept that doesn’t exist in the current repo

Usage

tribal recall --all "your search query"

Results include the assistant name (project identifier) alongside each memory, so you know where the knowledge came from.

Examples

# Find auth-related knowledge across all projects
tribal recall --all "JWT token refresh"

# Search for deployment patterns
tribal recall --all "Docker compose production setup"

# JSON output for agents
tribal recall --all --json "database migration rollback"

Agent behavior

When you set up agent integration with tribal setup-agents, agents are instructed to:

1. First try a local recall (current project only)
2. If no relevant results are found, automatically retry with --all

This two-step approach keeps searches fast by default while ensuring knowledge isn’t missed.

How it works

Each project gets its own Backboard assistant (created during tribal init). The --all flag tells Backboard to search across all assistants tied to your API key. Results are ranked by semantic similarity regardless of which assistant they belong to.

Considerations

• Cross-repo search is slightly slower than local search (more vectors to scan)
• Results from other projects may lack local context — read them as hints, not prescriptions
• All projects under the same Backboard API key are searchable — if you need isolation between teams, use separate API keys