Documentation

Lore docs.

Install Lore, bring your own provider key, initialize local memory, control permissions, verify changes, and run headless agent tasks.

Fast path: clone Lore, build the CLI, set a provider key, run lore init, then start lore. There is no Lore-hosted model gateway.

Installation

Lore is an open-source Go CLI. The module path is github.com/fayzkk889/lore. Build it from the public repository:

git clone https://github.com/fayzkk889/lore
cd lore
go build -o lore .

Release binaries for v0.9.1-beta will be available from GitHub Releases once published.

Release installers

Installer scripts can be used for release builds when artifacts are available:

curl -fsSL https://raw.githubusercontent.com/fayzkk889/lore/main/install.sh | bash
irm https://raw.githubusercontent.com/fayzkk889/lore/main/install.ps1 | iex

Quickstart

export OPENROUTER_API_KEY=sk-or-v1-...
lore init
lore

Configure defaults when you prefer not to rely on environment variables:

lore config set
lore config permission auto-safe
lore --provider ollama --model qwen3:4b

Providers

Lore connects directly from your machine to the model provider you choose. Supported paths include Anthropic, OpenAI-compatible providers, OpenRouter, DeepSeek, custom endpoints, and local Ollama.

export ANTHROPIC_API_KEY=sk-ant-...
export OPENAI_API_KEY=sk-...
export OPENROUTER_API_KEY=sk-or-v1-...
export DEEPSEEK_API_KEY=sk-...

For custom endpoints, configure the provider and base URL with lore config set. For local models, keep Ollama running and use --provider ollama --model qwen3:4b.

Wiki memory

Lore stores project knowledge locally in .lore/. The wiki includes architecture, conventions, logs, memory, generated pages, snapshots, and verification runs.

.lore/
  index.md
  log.md
  memory.md
  schema.md
  conventions.md
  architecture/
    overview.md
  learnings/
  snapshots/
  runs/

Add LORE.md to your project root for standing instructions: test commands, coding rules, architecture constraints, and review habits.

Permissions

Lore supports four local safety modes:

  • full-auto: fastest mode for trusted repositories.
  • auto-safe: automated work with more cautious defaults.
  • ask: prompt before sensitive actions in the interactive TUI.
  • read-only: analysis and review without file edits.
/permissions ask
/permissions read-only
lore config permission auto-safe
lore do --permission read-only "review this project"

Verified completion

Lore can verify applied work with build, vet, test, and runtime checks. Verification output is recorded under .lore/runs so you can inspect what was actually run.

/status
/runs
/audit

Rollback snapshots

Before applying changes, Lore creates local snapshots. If a change goes wrong, restore from snapshot history.

lore history
lore rollback
lore rollback --last

Snapshot hardening avoids obvious secret material and preserves skipped existing files during rollback.

Commands

CommandPurpose
loreStart interactive session
lore initScan project and create wiki memory
lore do "task"Run a headless agent task
lore config setConfigure provider and model defaults
lore config permissionShow or set default safety mode
lore historyList rollback snapshots
lore rollbackRestore a snapshot

TUI commands

/memory
/remember ship the API through the gateway package
/wiki
/recall auth middleware
/status
/runs
/audit
/rollback
/permissions auto-safe

Configuration

Lore stores user config in ~/.lore/config.toml.

[provider]
name = "openrouter"
model = "anthropic/claude-3.5-sonnet"

[safety]
permission_mode = "auto-safe"

[selfcheck]
enabled = true
Privacy note: project memory, snapshots, audit logs, and verification ledgers are local files. Review .lore/ before sharing a repository publicly.