phren

Fresh stores enable memory only. Enable what you need: modules add tasks, phone Hook, Git, schedules, conductor and the optional @phren/code package with phren modules enable|disable <name>.

Install

npx @phren/cli init

One command. Sets up ~/.phren, wires up MCP for your tools, installs hooks. Next time you open a project, context starts flowing automatically. On a new machine? Re-run init and you're back in sync.

Periodic remote checks are off by default. Enable them with phren config pull-interval 60 to check shared Git stores every minute while an MCP server is running. Clients sharing a store coordinate one check; only changed remote refs trigger a fetch, and automatic updates are fast-forward only. Use phren config pull-interval 600 for ten minutes, or phren config pull-interval off to disable periodic checks. See Git sync for details.

Related MCP server: MCP Memento

What actually happens

When you open a prompt:

• Hooks extract keywords from your question

• Phren searches findings across projects (FTS5 full-text with semantic fallback)

• Relevant snippets inject into your prompt before you hit send

• You ask; Claude already knows the gotchas

When you discover something:

• phren add-finding <project> "finding text" captures it with optional tags ([decision], [pattern], [pitfall], [bug])

• Trust scores decay over time; decisions never do; observations expire in 14 days

• Findings link to fragments (named concepts like "auth" or "build") that connect knowledge across projects

Sessions:

• Mark boundaries with session_start / session_end

• Next session sees your prior summary, active tasks, recent findings, and where you left off

• Checkpoints track edited files and failing tests so you can resume exactly where you stopped

Tasks:

• Add with priority/section. Pin across sessions. Link to GitHub issues.

• Track completions and cross-project rollups.

Interfaces

One store, seven ways in. Every surface reads and writes the same markdown files, so nothing you do in one is invisible to the others.

The shell opens with a short splash: the phren mascot beside the wordmark, which is revealed with a decrypt text effect on the first launch of a new version and shimmers while it waits for a key.

Watch it work

Put the graph in one terminal and an agent in another. Every memory a search lands on, every memory a hook injects, and every finding written is appended to a log the graph tails: the node pulses, the camera flies to it, the finding's full text fills the pane, and the event joins an activity feed. Writes show in green so saving is as visible as reading.

With PHREN_FEATURE_AGENTS=1 the graph also shows who is doing it. phren asks whatever is already running your agents, a Herdr workspace, phren-agent --multi, and joins each one onto a project by the directory it is working in. Tab cycles them, ↵ brings one to the front in its own host. Any tool that can print a small JSON record is a provider, so tmux or Zellij users need a few lines of shell rather than a change to phren.

Install it as a Claude Code plugin

If you would rather install phren the way you install everything else:

/plugin marketplace add alaarab/phren
/plugin install phren@phren

That brings the memory hooks (context at session start, recall on every prompt, save at stop), the MCP server and six phren-* skills, all version-pinned and removable with /plugin uninstall. On first run Claude offers to create the store with phren init. Plugin and phren init can coexist: the plugin's hooks yield to matching hooks from init, and its MCP server yields when Claude Code loads the server registered by init. See docs/claude-code-plugin.md.

phren also has its own coding agent. npm install -g @phren/agent, then phren agent -i: it starts every session already knowing the project's gotchas, tasks, and decisions, saves what it learns back to the same store, and runs on a ChatGPT/Codex subscription, OpenAI, OpenRouter, Anthropic, or local Ollama models. See docs/agent.md.

Key features

Fragment graph

Explore connections visually, in the browser or in the terminal. Drag nodes to reorganize in the web UI; the terminal layout is deterministic so the same store always draws the same map. Click a fragment to see every finding linked to it across all projects. The terminal graph also draws edges the browser does not: fragments co-mentioned by the same documents, and supersedes / contradicts links between findings.

Finding lifecycle

• Supersede: "Finding X is obsoleted by finding Y"

• Retract: "We were wrong about this; here's why"

• Contradict: "We have two findings that conflict; this is why"

Helps you reason about contradictions instead of hiding them.

Multi-agent support

Same store works with Claude Code, Copilot, Cursor, and Codex. Agents tag findings with their tool, so you see who discovered what.

Review queue

Mark findings as needing review ([Review] section). Phren surfaces review items on every session start. Approve, reject, or edit in place.

Governance & policies

Per-project retention policies. Confidence decay curves. Access control. Audit logs. Configure with phren config or the web UI.

Store subscriptions

Subscribe to specific projects in a team store; others stay hidden from search and context injection:

phren store subscribe team-store arc intranet
phren store unsubscribe team-store legacy-projects

Progressive disclosure

Enable PHREN_FEATURE_PROGRESSIVE_DISCLOSURE=1 to get compact memory indices instead of full snippets. Call get_memory_detail(id) to expand only what you need.

Semantic dedup & conflict detection

Optional: enable LLM-based duplicate detection and contradiction flagging on add_finding. Prevents near-duplicate entries and catches "always use X" vs "never use X" contradictions.

Skills & hooks

Drop custom slash commands into ~/.phren/global/skills/. Hooks run on user prompt, tool use, and session events; wire phren into your own workflows.

Herdr plugin

Working inside Herdr? herdr plugin install alaarab/phren/integrations/herdr binds the dashboard to a key: tasks, findings, and the review queue for whatever project the pane is sitting in, popped over your layout and gone again when you close it. See integrations/herdr.

Scheduled prompts

A project's schedules.yaml runs an agent on an assigned computer at a local time (interval, daily, weekly, once, or cron), launched by that computer's Phren Hook through Herdr or headless (docs/schedules.md).

Conductor

One conversation reads the project's tasks and sends independent briefs to enrolled computers with phren dispatch, which places a first prompt over pinned SSH and returns a durable receipt instead of a completion claim (docs/conductor.md).

iOS app

phren for iOS is a closed-source app for project memory and running agents; its source is not in this repository. GitHub sync keeps findings, notes, tasks and skills on your phone. Phren Hook adds Codex, Claude Code, Copilot and OpenCode chat, terminals, repository changes, the Code index and local app previews over pinned SSH/Tailscale. Steer working agents, remember notes about a function and send them to its agent, and reach a conductor through Siri or the Action button. Memory offers map and list views; local schedule and approval notifications need no relay. Set up Phren Hook on each computer with npx --yes @phren/cli@0.3.29 bridge install.

CLI quick reference

phren has 75 registered commands: 56 you will use (setup, projects, core, skills, hooks, config, maintain, stores, team) and 19 internal ones that hooks and background jobs call. phren --help prints the cheat sheet; phren help <command> the details.

phren                                   Interactive memory shell
phren search <query>                    Full-text search with FTS5
phren add-finding <project> "insight"
phren note add <project> "daily note"  Add a lightweight daily note
phren task add <project> "item"         Add a task
phren session_start <project>           Start a session
phren store list                        List personal + team stores
phren team init <name> --remote <url>
phren team join <url>                   Join a team store
phren shell --view tasks --here         Open the shell on this project's tasks
phren shell --view graph                Walk the knowledge graph in the terminal
phren web-ui [--port 3499]              Launch the web UI (3D graph, dashboard)
phren doctor                            Health check & auto-fix

See full CLI docs at alaarab.github.io/phren.

Team stores

Shared knowledge repos for teams. One person creates with phren team init, others join with phren team join. Findings, daily notes, tasks, and skills sync across team members.

Each team store can be configured with per-project subscriptions so people only see what they care about.

Platforms

Agents that write to the store:

• Claude Code (VS Code, Web, Desktop), MCP hooks + CLI

• Copilot (VS Code, GitHub.com), MCP hooks

• Cursor (IDE), MCP hooks + built-in skill system

• Codex, MCP tools + hooks

Ways to read it: the terminal shell and graph, the web UI, the VS Code extension, the iOS app, and the Herdr plugin (see Interfaces).

All use the same phren store. No vendor lock-in.

Packages

MIT License. Made by Ala Arab.