project-kb
Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type
@followed by the MCP server name and your instructions, e.g., "@project-kbsearch the latest source snapshot for 'calibration' and show its provenance"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
project-kb 1.0.1
A local, evidence-first project knowledge base with a Windows MCP server and CLI. It stores source snapshots and summaries in SQLite, retains provenance, and exposes bounded local search and exact readback. This public distribution has no external model API integration.
This source release contains code only. Runtime databases, vault content, local settings, credentials, logs, and source projects belong outside Git history. The production schema-3 deployment is Windows-specific. This release was tested on Python 3.11. Read the 1.0.1 audit and release notes before deploying it.
Try It With An Isolated Sample
Run these commands from a PowerShell session in this checkout. They create only
ignored .local data. Use a disposable sample project inside .local\sources.
The ranking byte ceiling is an explicit local storage limit, not a usage target.
py -3.11 -m venv .venv
& .\.venv\Scripts\python.exe -m pip install -r requirements-dev.txt
& .\.venv\Scripts\python.exe -m pytest -q tests
$repo = (Get-Location).Path
$env:PROJECT_KB_PROFILE = 'sandbox'
$env:PROJECT_KB_SANDBOX_BASE = $repo
$env:PROJECT_KB_ROOT = Join-Path $repo '.local\kb'
$env:PROJECT_KB_SOURCE_ROOT = Join-Path $repo '.local\sources'
$env:PROJECT_KB_PUBLIC_V2 = '1'
$env:PROJECT_KB_ENABLE_SANDBOX_WRITES = '1'
$env:PROJECT_KB_RANK_MAX_BYTES = '67108864'
New-Item -ItemType Directory -Path "$env:PROJECT_KB_ROOT\indexes", "$env:PROJECT_KB_SOURCE_ROOT\sample" -Force | Out-Null
Set-Content -LiteralPath "$env:PROJECT_KB_SOURCE_ROOT\sample\README.md" -Value 'A sample calibration checklist.'
& .\.venv\Scripts\python.exe scripts\kb_v2.py init
& .\.venv\Scripts\python.exe scripts\kb_v2.py init-api --operation-id sample-init-api
& .\.venv\Scripts\python.exe scripts\kb_v2.py register-root --operation-id sample-register --project-id sample --root "$env:PROJECT_KB_SOURCE_ROOT\sample"
& .\.venv\Scripts\python.exe scripts\kb_v2.py rank-install --operation-id sample-rank
& .\.venv\Scripts\python.exe scripts\kb_v2.py refresh --operation-id sample-refresh --project-id sample --root "$env:PROJECT_KB_SOURCE_ROOT\sample"
& .\.venv\Scripts\python.exe scripts\kb_v2.py search calibration --project-id sampleOperation IDs are idempotency keys. Reuse one only for an identical retry.
Keep PROJECT_KB_ENABLE_SANDBOX_WRITES unset for read-only sessions. The sample
does not turn a new checkout into an existing production deployment.
Start the MCP server from the same configured environment with
& .\.venv\Scripts\python.exe -B -X utf8 -u .\mcp_server.py.
The server exposes compact kb_retrieve and exact kb2_* interfaces; source
summaries remain evidence candidates until independently checked.
Related MCP server: agent-fact-system
Existing Production Store
An existing schema-3 installation has a deployment.json marker and an active
indexes/kb-v3.sqlite. By default, production code expects that installation at
$HOME\project-kb and source repositories at $HOME\git. Set
PROJECT_KB_INSTALL_ROOT and PROJECT_KB_SOURCE_ROOT before starting the server
when those paths differ. Keep the production data and code together at the
configured installation root. See local settings.
Search remains local. No query or indexed evidence is sent to an external model
service by this code. The local MCP interface and versioned kb2_* commands
remain available.
Before GitHub
Run python scripts\check_public_tree.py and inspect git status --short.
The check examines both Git's staged bytes and current worktree files. It
reports possible leaks by path and line number without printing matched text.
It cannot prove that every project document is licensed for publication.
After review, run python scripts\build_release.py to create the source ZIP
under the ignored .local\dist directory. The build refuses mismatched staged
and worktree bytes; it does not commit or push the repository.
License
This is source-available software, not an open-source license. Individuals and organizations may download, install, and run the unmodified code for their own internal purposes, including commercial internal use. Modification, redistribution, and public hosting require separate written permission. Read the complete LICENSE.md before use. GitHub's platform terms still allow viewing and forking a public repository.
The release does not upload anything automatically. Review and test details are in RELEASE_1.0.1.md. See SECURITY.md for the SSH, MCP and data boundaries.
This server cannot be deployed
Maintenance
Related MCP Connectors
- memoricOAuthio.memoric
Provenance-first database for teams and agents: every value carries sources, rules and coverage.
Machine-readable entity discovery with provenance, trust and verified source evidence.
Read-only game, setup, place, evidence and travel decision tools with explicit provenance.
Agent memory that refuses to guess: evidence-gated recall, exact-source reads, verifiable deletion.
Related MCP Servers
- AlicenseAqualityBmaintenanceProvides a compiled knowledge substrate, extracting atomic claims from an append-only capture log and querying a bitemporal claim graph over MCP. Read-only by default, with opt-in writes for trusted sources.4MIT
- AlicenseBqualityCmaintenanceLocal-first knowledge system for reasoning agents, exposing facts, evidence, documents, retrieval, and audit history through a thin stdio MCP server.14MIT
- FlicenseDqualityBmaintenanceProvides a local SQLite-backed code context knowledge base with MCP tools for storing and querying code facts, call graphs, semantic info, evidence, and business mappings, plus versioned snapshot publishing and incremental sync.776-
- AlicenseAqualityBmaintenanceEnables users to create frozen, auditable snapshots of selected Markdown or text-layer PDFs and perform read-only, verifiable local searches with SQLite FTS5/BM25, preserving source hashes, schemas, and page/line anchors.82Apache 2.0