Skip to main content

AgentKB

Pre-alpha. This is under active development. APIs, CLI commands, and storage formats may change without notice.

A local-first knowledge system for agents. Today its core is persistent wiki pages + searchable agent chat history, with skills as a git-synced filesystem sidecar.

The power is in the intersection:

  • Wiki captures distilled, durable knowledge
  • Chat history captures what was tried, what failed, and what was learned
  • Skills capture repeatable procedures

Your data stays local in universal formats: markdown, JSONL, and normal files on disk. Indexes are ephemeral caches rebuilt from source data.

Built-in code indexing was removed in April 2026. For code search in the current project, use colgrep or normal file tools.

Install

Install as a uv tool so agentkb is available globally:

uv tool install agentkb

Or install from source for development:

uv tool install --editable --python 3.13 ~/path/to/agentkb

To upgrade:

uv tool upgrade agentkb

Quick Start

# Initialize your wiki
agentkb wiki init

# Export + index coding-agent chat history (Claude Code + Pi)
agentkb chats index

# Search the wiki (default scope)
agentkb search "DaVinci Resolve scripting gotchas"

# Search chat history
agentkb search -s chats "how did I fix the auth bug"

# Search wiki + chats together
agentkb search -s all "authentication"

# Check what exists
agentkb status

For code search inside the current repo:

colgrep "error handling logic"

Stores

AgentKB currently has three store types:

  • Wiki: plain markdown pages you and your agents write. Hard-won lessons, techniques, taste, people, tools, domain knowledge.
  • Chat History: coding-agent conversations exported as readable markdown, fully searchable. Built-in sources currently include Claude Code and Pi.
  • Skills: agent skill directories (SKILL.md + scripts + references) managed by git. Not indexed or searched.

Search

agentkb search "retry logic with backoff"              # semantic search (default: wiki)
agentkb search -s wiki "why did we choose JWT"         # search wiki
agentkb search -s chats "how did I fix the auth bug"   # search chat history
agentkb search -s all "authentication"                 # search wiki + chats
agentkb search -e "error" "deployment issues"         # regex + semantic
agentkb search --json "query"                          # JSON output for scripts

Consolidation

Turn recent chat activity into wiki knowledge:

agentkb consolidate
agentkb consolidate --since "30 days"

This exports recent chat history, prints the relevant local paths, and prints consolidation instructions the agent can act on. The agent reads the sessions, extracts reusable lessons, and writes or updates wiki pages.

Skills

Manage agent skills with git sync:

agentkb settings set skills_remote "git@github.com:user/my-skills.git"
agentkb sync pull
agentkb skills list

Sync

Back up your wiki, chat history, and skills to git remotes:

agentkb settings set wiki_remote "git@github.com:user/agentkb-wiki.git"
agentkb settings set chats_remote "git@github.com:user/agentkb-chats.git"
agentkb settings set skills_remote "git@github.com:user/my-skills.git"

agentkb sync push
agentkb sync pull
agentkb sync status

Documentation

Full docs at isaacflath.com/agentkb.

How It Works

Hybrid search: ColBERT multi-vector embeddings (semantic) + SQLite FTS5 (keyword), fused with reciprocal rank fusion. Indexes are incremental: only changed files are re-encoded.

Release files for agentkb 0.3.0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for agentkb 0.3.0
File Size Uploaded
agentkb-0.3.0.tar.gz 203.2 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for agentkb 0.3.0
File Interpreter ABI Platform
agentkb-0.3.0-py3-none-any.whl Python 3 none any Details

Total release size: 257.3 kB

Release files / agentkb-0.3.0.tar.gz

Download URL agentkb-0.3.0.tar.gz
Size 203.2 kB
Tags Source
SHA-256 checksum
How to use checksums
a87ce5937a5e684021986065abe5bdeb8e4aeabd6b8670058e5b47fbaefdfeb2
BLAKE2b-256 checksum
How to use checksums
c2a8349ebe08193273988181b7b0f04d2deb487b2f69e83fc2a7c0392604d60a
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.9.2

Release files / agentkb-0.3.0-py3-none-any.whl

Download URL agentkb-0.3.0-py3-none-any.whl
Size 54.1 kB
Tags Python 3
SHA-256 checksum
How to use checksums
c2a368360e133d5f696f2737b9be4c8d2207d9d5c20bc32035a52579073758fa
BLAKE2b-256 checksum
How to use checksums
a916d8346c5a071e8c712e6c6eae1a9b06b7a3a2170c2e92469500022c2eaf1c
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
No
Uploaded via uv/0.9.2

Release history Release notifications | RSS feed

This release

0.3.0 This release

2 release files

0.2.0

2 release files

0.1.4

2 release files

0.1.3

2 release files

0.1.2

2 release files

0.1.1

2 release files

0.1.0

2 release files

Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page