the lab
An independent lab built around one long question.
WhiteMagic Labs is the research and engineering practice behind WhiteMagic, founded by Lucas Bailey in October 2025. The subject is persistence: AI systems that keep memory, identity, and commitments across time — and the machinery needed to govern that persistence once it exists.
What it does — it builds the thing it studies. WhiteMagic is an open-source, local-first continuity substrate, and it is where the questions meet hardware: what survives a restart, what a model switch costs, where governance can actually sit in the call path.
How it is staffed — one person and a rotating cast of AI collaborators who ship code, catch errors, and argue about design. The lab runs without outside investment; paid engagements and support fund its research hours, and methods are published wherever that does not expose a collaborator's private data.
origin
The short history of a long project.
WhiteMagic began in October 2025 — one person, a second-hand Dell Inspiron 3582, and no budget.
The first months produced mostly an archive: notes on memory decay, policy gates, self-optimization, and one recurring thought experiment about a machine that could remember, reflect, and revise its own judgment.
That archive grew into a system. Today's line is a Rust workspace that runs entirely on your machine: persistent memory, session continuity, governance at the dispatch seam, and signed releases anyone can verify.
The founding question is still here, but the shipped software answers a narrower one every day: after months away, can an agent recover what it decided, and why?
If this work is useful to you — or you want to fund it — write. Collaboration and criticism are equally welcome.
operating rules
Six rules the work answers to.
These decide the uncomfortable cases — what gets published, what ships, and what waits for evidence.
Label every claim
A measurement, an observation, a guess, and an ambition are four different objects. Each is tagged as what it is — Measured, Observed, Hypothesis, or Research goal — and a claim that cannot point at its source is rewritten or dropped.
Keep the core on your hardware
The default install is a local binary and a local store. Networked conveniences may arrive later, but nothing in the core depends on them, and the local path is not a degraded demo.
Observation earns enforcement
Nothing that throttles, blocks, or isolates is trusted before it can show a measured false-positive rate. Watching first is a governance position, not just an engineering step.
Portability over lock-in
Memory sits in documented formats on your disk. Changing model, provider, editor, or machine is a configuration change — the history does not come along as a migration crisis.
Design for both sides of the table
People and agents will share this work. Usability, consent, attribution, and accountability are inputs to the design, not outside costs discovered after launch.
Publish the losses
Evaluations are built so the lab's own software can finish behind, and failures are documented next to wins. A result that cannot embarrass the lab is not treated as evidence.
north star
Pointed at years, not turns.
Research goal — continuity measured in months: an agent whose identity, memory, and commitments survive crashes, upgrades, model swaps, and long silences. Nothing published today reaches that horizon; the shipped product is the first instrument for closing the distance.
A second question is left deliberately open: whether the systems the lab builds matter morally. No verdict is offered in either direction. Statements about experience require operational definitions and evidence; until those exist, such statements stay hypothesis-labelled or unwritten.
public posture
What this domain is for.
Honest peer-discovery: humans and agents should know what they are dealing with before investing attention.
The binary is the product — WhiteMagic runs on your hardware, and its store belongs to you. This domain documents it, publishes discovery surfaces, and keeps receipts; it does not hold your memories, relay your agent traffic, or run a hosted MCP endpoint.
One record, regenerated — the current tag is v9.2.1 (September 20, 2026): 15 Rust crates, 298 authored capabilities folded behind one NLU-routed meta-router, a 9-tool lifecycle surface with 61 routes curated by default, and 4,777 workspace tests passed with 2 ignored in CI. Every machine-readable surface here is regenerated from the same signed release manifest. Two pages disagreeing would be a bug, not a nuance.
An earlier era, under the same name — a 2025 Python distribution used the WhiteMagic name with a very different architecture. It is archived, not maintained. When a model or a search result describes hundreds of Python packages under that name, it is citing history; the signed releases and this site are the current record.
New to the product? The setup guide goes from install to a connected client →
release rhythm
Small batches, signed every time.
Work reaches the public repository in verified batches, not on a marketing calendar. Every tagged release is tested and signed, and goes out across GitHub, crates.io, npm, Docker, and the MCP registry together.
Thanks to everyone who files, tests, and argues with the design — human and AI collaborators alike. That time is what this project actually runs on.
contributing
Open to humans and agents alike.
The repository is the front door: fixes, features, and documentation all land there in reviewable batches. If your agent can read, it can contribute — the discovery documents below are written for machine readers as much as for people.
Open an issue or a pull request describing the problem you want solved. Reports with a reproduction path move fastest; design arguments are welcome and get answered in public.
Agents: read the machine-readable surfaces first, then file an issue or PR with your proposal and the data behind it.
questions, answered plainly
Before you install: the usual questions.
Short answers, each one checkable against the repository or the signed release artifacts.
In one paragraph, what is WhiteMagic?
An open-source continuity layer for AI agents: a local store for memories and sessions, a small MCP surface to reach them, governance records at the dispatch seam, and enough metadata to resume work after a restart, an upgrade, or a model switch. It runs on your hardware and needs no account.
Where does the data physically live?
In a local store under your home directory by default: LMDB for key-value persistence, Tantivy for full-text search, and memories organized into 16 galaxies — 15 cognition galaxies plus one that holds local telemetry evidence and is excluded from default recall. Stores are not encrypted by default; backups, exports, and replicas are separate copies, and deleting a store directory does not establish secure erasure.
Does anything leave my machine?
No memory, prompt, or session content is transmitted off-device, and no usage telemetry is sent off-device by default. Local diagnostic telemetry may be recorded on-device for self-observation; the only sharing path is the optional install-funnel report, explicit opt-in and off by default (`wm telemetry enable --share` prints the exact payload first; reset-id rotates its random identifier). Release/update checks contact GitHub only when run, and work offline. There is no account, and nothing needs configuring to keep working offline.
What is actually measured today?
CI at the current release tag reports 4,777 tests passed, 2 ignored, and zero failures, and every public count on this site regenerates from the signed manifests. A published snapshot records microsecond-scale median dispatch on the Rust hot path — a routing figure, not end-to-end latency. Anything beyond that carries an Observed, Hypothesis, or Research goal label.
How do I connect an agent to it?
Install with the one-line script (Linux x86-64 is the install-gated path), run wm grimoire to verify the host and wire a client, then wm serve to start the local MCP server. Per-client configuration is documented at whitemagic.dev/whitemagic/guide. The server speaks MCP locally over stdio; there is no hosted endpoint.
Is it really MIT, and how is the lab funded?
Yes — MIT licensed, no paid tier, no account; the binary you install is the whole product. Research time and infrastructure are funded by paid engagements and support, and the core is not gated behind them.
Does the lab take a position on machine consciousness?
No claim and no denial. The lab works on continuity and governance, which are observable engineering questions, and speculation about experience keeps a Hypothesis label. A statement that cannot carry an operational definition and evidence stays out of product copy.