Skip to main content

CLI-first personal knowledge base for AI agents — structured, navigable, zero extra cost

Project description

kvault

Tell your AI agent to build you a knowledge base. That's it.

pip install knowledgevault

kvault gives your coding agent persistent, structured memory. It runs as a CLI tool that any agent can call via shell: terminal coding agents, editors, IDE assistants, or any tool that can execute commands. kvault itself requires no extra API keys, hosted service, or database.

Your agent creates nodes (people, projects, notes, categories), navigates the hierarchy via parent summaries, and keeps everything in sync — all through simple CLI commands.

Who is this for?

Developers using AI coding tools who want their agent to remember things between sessions — contacts, projects, meeting notes, research — in a structured, navigable format.

What makes it different?

kvault Flat memory files Hosted docs / memory apps Note-vault templates
Structure Hierarchical nodes with navigable tree Flat JSON Rich docs, flat search Obsidian vault
Agent-native CLI commands, works in any subprocess MCP server Chat/sidebar workflow Template, not runtime
Service model Plain files + local CLI Local server Hosted workspace Local vault
Navigation Parent summaries at every level None AI-generated Manual
Search Structured node search + native grep/find Built-in Built-in Manual

Quickstart (30 seconds)

1. Install

pip install knowledgevault

2. Initialize a knowledge base

kvault init ./my_kb --name "Your Name"

3. Tell your agent

"Use kvault CLI commands to manage my knowledge base at ./my_kb"

Your agent reads the generated AGENTS.md for workflow instructions and starts working. Use --kb-root ./my_kb from the directory containing my_kb, or pass an absolute path.

Tool setup tips:

Tool Setup
Project-instruction agents Keep AGENTS.md in the KB root so the agent reads the workflow automatically
Terminal agents Tell the agent: "Read AGENTS.md for the kvault workflow, then use shell commands to manage ./my_kb"
Custom-instruction agents Paste the generated AGENTS.md workflow into the workspace or system instructions

Try it: import an exported history

The best way to see kvault in action is to point it at data you already have. Many chat, email, calendar, and notes tools can export JSON, Markdown, CSV, mbox, or zip archives. Your agent can turn those raw exports into a structured, navigable knowledge base in minutes.

Exports can contain sensitive information. kvault stores files locally in your KB, but your agent may read excerpts while processing them and send those excerpts to its model provider. Review or redact exports first if that matters for your use case.

1. Export source data

Download an archive from a tool you already use. Keep the raw files under sources/ so they stay separate from curated nodes.

2. Unzip it into your KB

mkdir -p my_kb/sources/conversations
unzip conversation-export.zip -d my_kb/sources/conversations

3. Tell your agent to process it

Read through the exported files in sources/conversations.
Extract the people, projects, and ideas I've discussed most frequently.
Create nodes for each one in the knowledge base.

Your agent will use kvault CLI commands to create structured nodes with frontmatter and propagate summaries.

The CLI 2-call write workflow

# Call 1: Write node (stdin = frontmatter + markdown body)
kvault write people/contacts/acme --create --reasoning "New customer" --json --kb-root ./my_kb <<'EOF'
---
source: meeting_2026-02-25
aliases: [ACME Corp]
---
# ACME Corp
Key customer acquired at trade show...
EOF
# → {"success": true, "ancestors": [{path, current_content, has_meta}, ...]}

# Call 2: Agent reads ancestors, composes updated summaries, including root
kvault update-summaries --json --kb-root ./my_kb <<'EOF'
[
  {"path": "people/contacts", "content": "# Contacts\n...updated..."},
  {"path": "people", "content": "# People\n...updated..."},
  {"path": ".", "content": "# Knowledge Base\n...updated..."}
]
EOF
# → {"success": true, "updated": ["people/contacts", "people", "."], "count": 3}

What a node looks like

Each node is a directory with a single _summary.md file containing YAML frontmatter. Leaf nodes often represent entities such as people, projects, or notes; parent nodes summarize their descendants.

---
created: 2026-02-06
updated: 2026-02-06
source: manual
aliases: [Morgan Lee, morgan@example.com]
email: morgan@example.com
---
# Morgan Lee

Research collaborator tracking evaluation notes and follow-up questions.

Required frontmatter: source, aliases (kvault sets created/updated automatically)

What a knowledge base looks like

my_kb/
├── _summary.md                          # Root: executive overview
├── AGENTS.md                            # Agent workflow instructions
├── people/
│   ├── _summary.md                      # "12 contacts across 3 categories"
│   ├── family/
│   │   └── _summary.md
│   ├── friends/
│   │   ├── _summary.md
│   │   └── alex_rivera/
│   │       └── _summary.md
│   └── contacts/
│       ├── _summary.md
│       └── sarah_chen/
│           └── _summary.md
├── projects/
│   ├── _summary.md
│   └── launch_plan/
│       └── _summary.md
├── journal/
│   └── 2026-02/
│       └── log.md
└── .kvault/
    └── logs.db                          # Observability

CLI commands

Category Commands
Node kvault search, kvault read, kvault write, kvault list
Compatibility kvault read-summary, kvault write-summary, kvault ancestors, kvault delete, kvault move
Journal kvault journal
Status kvault status, kvault tree
Validation kvault validate, kvault check
Artifacts kvault artifact daily
Logs kvault log summary
Init kvault init

Agent-facing commands support --json for machine-readable output. --kb-root overrides auto-detection on KB-bound commands, and it works before or after the subcommand:

kvault read people/friends/alice --json --kb-root ./my_kb
kvault search "alice project notes" --json --kb-root ./my_kb
kvault artifact daily --json --kb-root ./my_kb

kvault search is structured node discovery: it searches visible _summary.md files, ranks title, path, alias, heading, and body matches, and returns node paths with snippets. Raw filesystem search is still useful for exact text investigation:

rg -n "alice|project" ./my_kb
kvault search "alice project" --json --kb-root ./my_kb
kvault read people/friends/alice --json --kb-root ./my_kb

kvault read <path> is canonical context retrieval. It returns the full requested node plus its immediate parent summary by default.

File-native browsing

kvault is just Markdown files and CLI tools. For navigation, combine structured commands with native file search or your preferred Markdown viewer:

kvault tree --kb-root ./my_kb
kvault search "alice project" --json --kb-root ./my_kb
kvault read people/friends/alice --json --kb-root ./my_kb
rg -n "alice|project" ./my_kb

Optional MCP compatibility

CLI remains the primary interface for shell-capable agents. For MCP-native clients, kvault also ships a stdio MCP compatibility server:

pip install "knowledgevault[mcp]"
kvault-mcp --kb-root /absolute/path/to/my_kb

knowledgevault[mcp] requires Python 3.10+. The server is bound to one KB root per process; use a separate kvault-mcp process for each KB. You can pass the root with --kb-root or set KVAULT_KB_ROOT.

Generic MCP client config:

{
  "mcpServers": {
    "kvault": {
      "command": "kvault-mcp",
      "args": ["--kb-root", "/absolute/path/to/my_kb"]
    }
  }
}

Root-pinned config for shared runtimes:

{
  "mcpServers": {
    "kvault": {
      "command": "kvault-mcp",
      "args": ["--kb-root", "/absolute/path/to/my_kb"],
      "env": {
        "KVAULT_ALLOWED_ROOTS": "/absolute/path/to/my_kb"
      }
    }
  }
}

The compatibility server exposes root-bound tools for node search/read/write/list, legacy entity and summary calls, ancestor lookup, journaling, daily artifact generation, validation, and phase logging:

Category Tools
Lifecycle kvault_init, kvault_status
Nodes kvault_search, kvault_read_node, kvault_write_node, kvault_list_nodes
Compatibility kvault_read_entity, kvault_write_entity, kvault_list_entities, kvault_read_summary, kvault_write_summary, kvault_delete_entity, kvault_move_entity
Summaries kvault_prepare_summary_update, kvault_write_parent_summary, kvault_update_summaries, kvault_get_parent_summaries, kvault_get_ancestors, kvault_propagate_all
Journal / artifacts kvault_write_journal, kvault_generate_daily_artifact
Validation / logging kvault_validate_kb, kvault_log_phase

Compatibility tools accept an optional kg_root argument for older clients, but it must match the server-bound root. kvault_init reports bound-root status and rejects mismatched roots; it does not create or reinitialize a KB.

MCP clients should prefer the strict parent-summary workflow:

  1. Call kvault_status, kvault_list_nodes, or kvault_search to orient.
  2. Call kvault_read_node before editing; it returns the node plus parent context.
  3. Call kvault_write_node with Markdown body content and durable metadata in meta.
  4. For each returned ancestor, closest-first, call kvault_prepare_summary_update.
  5. Compose the parent summary from the returned parent and immediate child summaries.
  6. Call kvault_write_parent_summary with the new content and the returned children_digest.
  7. Call kvault_validate_kb after larger edits.

children_digest is stateless and only proves the direct child summaries have not changed since the prepare call. If another write changes a direct child first, kvault_write_parent_summary returns workflow_error; call kvault_prepare_summary_update again and compose from the current children.

When a parent has more than 10 direct children, kvault_prepare_summary_update returns a hierarchy_hint. This is advisory: split the hierarchy when natural groups are obvious, otherwise still keep the parent summary comprehensive. Compatibility tools such as kvault_update_summaries and kvault_write_summary remain available for older clients and manual maintenance.

Optional root pinning (multi-tenant hardening)

For shared runtimes, pin allowed roots:

export KVAULT_ALLOWED_ROOTS="/absolute/path/to/my_kb"

Python API

from pathlib import Path
from kvault.core import operations as ops

kg_root = Path("my_kb")

# Read/write/search nodes
node = ops.read_node(kg_root, "people/contacts/sarah_chen")
result = ops.write_node(kg_root, "people/contacts/new_person", "# Content", create=True)
matches = ops.search_nodes(kg_root, "sarah follow up")
prepared = ops.prepare_summary_update(kg_root, "people/contacts")
ops.write_parent_summary(
    kg_root,
    "people/contacts",
    "# Contacts\n\nUpdated rollup.",
    prepared["children_digest"],
)

# Entity reconciliation remains available for dedup decisions
from kvault import scan_entities, EntityResearcher
entities = scan_entities(kg_root)
researcher = EntityResearcher(kg_root)
action, target, confidence = researcher.suggest_action("Morgan Lee")

Integrity check

Run kvault check to catch stale summaries and weak parent rollups:

kvault check --kb-root /absolute/path/to/my_kb

[KB] warnings keep the existing nonzero exit behavior. SUMMARY: warnings are warn-only by default, capped at 5 lines, and call out parent summaries that are too short, omit immediate children, or contain placeholder/redirect language. Use --no-summary-quality to skip that audit.

If your tool supports pre-prompt hooks, you can automate this:

{
  "hooks": {
    "UserPromptSubmit": [
      {
        "type": "command",
        "command": "kvault check --kb-root /absolute/path/to/my_kb"
      }
    ]
  }
}

It's just files

kvault produces Markdown with YAML frontmatter in a plain directory. No proprietary format, no database to export from. Your existing tools work out of the box:

Want to... Use
Semantic search Embed the .md files with any vector tool
Exact text search rg -n "phrase" ./my_kb
Visual browsing Open the KB directory in Obsidian or Logseq
Publish as a site Point Hugo, Jekyll, or Astro at the directory
CI validation Run kvault validate or kvault check in a GitHub Action
Bulk export find . -name _summary.md + yq over the frontmatter

Development

pip install -e ".[dev,mcp]"
pytest -q
ruff check .
black --check kvault/ tests/
mypy kvault/ --ignore-missing-imports

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

knowledgevault-0.10.0.tar.gz (72.1 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

knowledgevault-0.10.0-py3-none-any.whl (61.0 kB view details)

Uploaded Python 3

File details

Details for the file knowledgevault-0.10.0.tar.gz.

File metadata

  • Download URL: knowledgevault-0.10.0.tar.gz
  • Upload date:
  • Size: 72.1 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for knowledgevault-0.10.0.tar.gz
Algorithm Hash digest
SHA256 13ebbedd56c7e54e8bc0a415a925bc6c73448b8f134b8c4f6c573b92c32dcf68
MD5 5a0225835f48791dfb678bc524ce4680
BLAKE2b-256 ccb5db087612b9ba99471c28db099383ce5a4d99807a75fa75242a52100d007e

See more details on using hashes here.

Provenance

The following attestation bundles were made for knowledgevault-0.10.0.tar.gz:

Publisher: publish.yml on cimo-labs/kvault

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file knowledgevault-0.10.0-py3-none-any.whl.

File metadata

File hashes

Hashes for knowledgevault-0.10.0-py3-none-any.whl
Algorithm Hash digest
SHA256 4ea54a4d449c9c2dbbd386be73310790f18270d6dc54403c25aec13248fb7364
MD5 946b13a87aa96f76a5b3ed8affdae829
BLAKE2b-256 0a45a61a3d00f28b3ef26ef321129e9ac97898dca519e89eceaf2f386c7be21a

See more details on using hashes here.

Provenance

The following attestation bundles were made for knowledgevault-0.10.0-py3-none-any.whl:

Publisher: publish.yml on cimo-labs/kvault

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page