Reference

Install

macOS (Apple Silicon) and Linux (x86_64):

curl -fsSL https://raw.githubusercontent.com/tasker-systems/temper/main/scripts/install/install.sh | sh

Windows (x86_64, PowerShell):

irm https://raw.githubusercontent.com/tasker-systems/temper/main/scripts/install/install.ps1 | iex

Supported platforms: macOS (Apple Silicon), Linux (x86_64), and Windows (x86_64). File issues at github.com/tasker-systems/temper/issues if you hit problems.

To pin a version, pass --version vX.Y.Z to the install script (… | sh -s -- --version vX.Y.Z). The install playbook covers Homebrew, uninstalling, other platforms, and troubleshooting.

Build from source

If you'd rather compile locally (Intel Mac, Linux arm64, or just to hack on temper itself):

git clone https://github.com/tasker-systems/temper.git
cd temper
cargo install --path crates/temper-cli --locked --features embed,extract

The embed feature pulls in ONNX Runtime for local embeddings; extract enables document ingestion via kreuzberg. Both are enabled by default.

Getting started

The local vault is a directory of markdown files with YAML frontmatter — a read-only projection of cloud state, not the source of truth (see Cloud). Temper resolves its config from ~/.config/temper/config.toml or a per-project .temper/config.toml.

To add content, create a resource and pipe in a body — temper resource create --type research --title "…" --body @notes.md. See Resources.

Resources

Every piece of knowledge in temper is a resource — a markdown file with typed frontmatter. The CLI's --type flag accepts six workflow types; the platform's full vocabulary is fourteen doc types — these six plus fact, memory, question, theme, concern, principle, commitment, and domain. The MCP server's describe_schema tool enumerates the live list. The six workflow types:

Resources are addressed by ref — a UUID, or the decorated slug-<uuid> form (resolved by the trailing UUID; the slug half is presentation, so a stale slug is harmless). Every list / show / search row prints a ref field — copy it, paste it.

Create · list · show · update · delete

Search

temper search combines full-text and semantic search — your query is embedded locally and matched by meaning, not just keywords.

Relationships

Resources aren't isolated — typed edges connect them, and search can expand along those edges to surface neighbors. Edges are first-class: asserted explicitly, then re-typed, re-weighted, or folded as the understanding changes. Source and target are refs; an edge is addressed by its own handle (a UUID).

Cloud

Temper Cloud is the Postgres-native source of truth, with pgvector-powered semantic search. The local vault is the inverse of the old model: a read-only projection cache, not the canonical copy. Writes (resource create/update/delete, edge) route straight to the API and take effect immediately; temper pull <ctx> re-materialises the local markdown from server state. Removing a projected file with rm is just a local cache miss — it has no server effect.

Auth

Teams

Agents

Temper is built to be legible to AI coding agents. Two integration paths — a Claude Code skill file, or the remote MCP server.

Skill file

A generated skill file teaches an agent your vault's structure, doc types, and workflow vocabulary — for Claude Code, opencode, or any ~/.agents-compatible tool:

temper skill install

To automatically prime new agent sessions with recent context, add a SessionStart hook:

{
  "hooks": {
    "SessionStart": [{
      "hooks": [{
        "type": "command",
        "command": "temper warmup --context @me/myapp"
      }]
    }]
  }
}

MCP server

The remote MCP server exposes vault operations as structured tools over Streamable HTTP. Agents authenticate via OAuth 2.1 + PKCE — you'll be prompted to log in on first connection. Connect Claude Desktop or Claude Code:

{
  "mcpServers": {
    "temper": {
      "url": "https://temperkb.io/mcp"
    }
  }
}

Core tools

The tool surface also covers cognitive maps (cogmap_*), blobs (blob_*), data artifacts (commit_data_artifact and friends), and facets (facet_set / facets_read) — the full list is what the server advertises over MCP, so your client's tool list is the reference.

Config

Global config lives at ~/.config/temper/config.toml. A per-project .temper/config.toml overrides it when temper is run from that directory.