Palinode

┌─ palinode ─┐

│ ░░░░░░░░░░ │ │ ▓▓▓▓▓▓▓▓▓▓ │ │ ██████████ │ └────────────┘

Inspectable, correctable project memory for repeated work across sessions, agents, and worktrees.

Last week, your team chose SQLite for a cache because it avoided operating a database. This week, a changed requirement makes PostgreSQL the better choice. Without the decision and its reason, a fresh agent can repeat the old approach. With Palinode, you can save the replacement, inspect the original and its history, and correct the record for the next session.

Palinode keeps project memory in Markdown files and Git history under your local control. An agent can use its tools to save, search, and inspect that memory. Some supported client integrations can also perform configured capture or recall; their automatic behavior, scope, and controls vary by client and configuration. Nothing is silently promoted from an unmarked note to a fact.

*A palinode is a poem that retracts what was said before and says it better. That's what memory compaction does.*

Built by Paul Kyle at phasespace-labs. See AUTHORS.

Start here

Follow the canonical Quickstart for one complete journey: install → start → connect an MCP client → save a decision → open a fresh session → inspect its markdown/provenance → correct it. It keeps the private memory store separate from the code checkout and distinguishes the current release from source-checkout capabilities.

Inspect the retired decision

The screenshot shows the archived initial SQLite decision after the separate PostgreSQL replacement was saved and the original was retired. See the local provenance UI guide for the full read-only inspector and the correction lifecycle; the guide explains that the image is taken after retirement, not before the save steps.

Supported Platforms

All platforms share the same MCP server — install once on your server, connect from any IDE. [docs/HARNESSES.md](docs/HARNESSES.md) is the cross-harness map: what each harness gets (native hooks vs. plugin vs. MCP), how the tiers stack, and where to start. Per-client config snippets: docs/MCP-SETUP.md and docs/MCP-INSTALL-RECIPES.md.

The Idea

Palinode treats plain files as the source of truth and builds its index and interfaces from those files.

Files (markdown + YAML frontmatter)

↓ watched Index (SQLite-vec vectors + FTS5 keywords, single .db file) ↓ queried by Interfaces (MCP server, REST API, CLI, OpenClaw plugin) ↓ compacted by Consolidation (structured operations → validation and application → git commits)

That's the whole architecture. One directory of .md files, one SQLite database, one API server. No Postgres, no Redis, no cloud dependency.

One Backend, Every Interface

Palinode doesn't care how you talk to it. The full toolkit — save, search, doctor, dedup-suggest, orphan-repair, diff, blame, rollback, and more — works through every interface:

Set up once on a server. Connect from any machine, any IDE, any agent framework. The MCP server is a pure HTTP client — it holds no state, no database connection, no embedder. Point it at the API and go.

{

"mcpServers": { "palinode": { "type": "http", "url": "http://your-server:6341/mcp/" } } }

That's the entire client config. Works with Claude Code, Claude Desktop, Cursor, Windsurf, Zed, and VS Code (Continue/Cline). palinode-mcp-http serves streamable-HTTP at /mcp/ — use "type": "http", not "type": "sse". Always include the trailing slash in the URL. See docs/MCP-SETUP.md for editor-specific install recipes.

How It Works

Store — Typed markdown files (people, projects, decisions, insights) with YAML frontmatter. Git-versioned. Human-readable. Editable in Obsidian, VS Code, vim, or anything.

Index — A file watcher indexes with FTS5 as you save. Content-hash dedup skips re-embedding unchanged files. The index is a local SQLite file; embeddings can use a local BGE-M3 endpoint or a configured remote provider.

Search — Hybrid BM25 + vector search merged with Reciprocal Rank Fusion. The two arms have different jobs: on full-sentence questions the vector arm can retrieve semantic matches while BM25 catches exact terms and identifiers. See benchmarks for the evaluation context and limitations. Optional associative entity graph and prospective triggers.

Compact — Weekly consolidation where an LLM returns structured operations and Palinode validates and applies them. Every compaction is a git commit you can review, blame, or revert.

Dream — If you've met "dreaming" as the name for this, palinode dream is an alias for palinode consolidate. Use --dry-run to inspect the proposed operations; each completed pass lands as a git commit, so a bad consolidation is a diff to review and a commit to revert.

Audit — git blame any fact. git diff any change. rollback any mistake. These aren't just git-compatible files — palinode_diff, palinode_blame, and palinode_rollback are first-class tools your agent can call.

Requirements

• Python 3.11+
• Git
• Ollama with bge-m3 (ollama pull bge-m3, ≈1.2 GB), or another supported embedding endpoint — for hybrid indexing and search. Since v0.21 you can instead choose explicit lexical mode; it is keyword/FTS retrieval, not a fallback when hybrid's endpoint fails. Saves persist without an embedder, but hybrid search returns HTTP 503 until it is reachable (palinode resolve degrades to keyword-only and says so). See the Homebrew setup guide for installation and verification.

Optional extras: a chat model for weekly consolidation (any 7B+ that outputs JSON), OpenClaw for agent plugin hooks.

Install

Before your first capture: Palinode stores readable Markdown and Git history. private/restricted control discovery by scope; a caller with API access can still read a hidden memory by its known path and use full-store maintenance tools. These labels provide no encryption or per-user/per-agent authentication. Protect the store, backups and API credentials; use separate instances or filesystem permissions for stronger separation. See the privacy contract.

The Quickstart is the authoritative first-use sequence. It covers the separate private store, second-terminal environment, supported lexical preview and hybrid model paths, editor connection, fresh-session check, inspector, and correction workflow. Do not combine a code checkout and a memory store. For a released Homebrew installation, begin with docs/HOMEBREW.md; for an autonomous agent bootstrap, see llms-install.md.

Running as a service

Three long-running processes (API, watcher, embedder) shouldn't live in terminal tabs. Pick one:

With compose, your memory stays on the host at ~/.palinode (override with PALINODE_DATA_DIR) — the containers mount it; files remain the source of truth. Already running Ollama on the host? OLLAMA_URL=http://host.docker.internal:11434 docker compose up -d palinode-api palinode-watcher skips the bundled one. Verify any of the three the same way: palinode doctor (or curl http://127.0.0.1:6340/status).

Connect your editor

Palinode speaks MCP. Follow the connection step in the Quickstart: palinode mcp-config generates a native fragment for the selected client, and the recipes keep the supported manual configuration. It is read-only; it never edits a client configuration file.

Connecting to a server on another machine over HTTP? Pass your project: palinode mcp-config --http --project adds an X-Palinode-Project header. Since v0.22 a remote server no longer scopes recall to its own working directory, so a client without the header gets unscoped results, reported as Scope: none. See the v0.22.0 compatibility notes. Full per-harness detail: docs/MCP-INSTALL-RECIPES.md.

Merge the Palinode entry into existing settings; redirecting generated output onto an existing configuration file would overwrite it.

Daily use — drop into a project

Already installed with palinode-api running? Scaffold any project in one command:

cd your-project

palinode init

That scaffolds .claude/CLAUDE.md (memory instructions, appended if one exists), .claude/settings.json (SessionStart + SessionEnd + UserPromptSubmit hook registration), all three hook scripts, and .mcp.json (points Claude Code at the palinode MCP server). Sessions then start smart, recall as they go, and end captured: the SessionStart hook injects your core: true memories into every fresh session (startup and /clear) so standing context is there before the first prompt; the UserPromptSubmit hook recalls relevant memory before each prompt — prospective triggers plus a strict-threshold search, injected as compact snippets, silent when nothing matches; and the SessionEnd hook auto-captures on /clear, logout, and exit. Re-run with --dry-run to preview, --force to overwrite, or --no-mcp / --no-hook to scope it. See examples/hooks/ for tuning knobs.

Projects that use other harnesses get the same memory instructions automatically: when AGENTS.md (or a .agent/ directory) exists, init appends a harness-neutral memory block to AGENTS.md (read by Codex, Antigravity, and other AGENTS.md-aware agents), and when a .cursor/ directory exists it writes `.curs