Personifypersonify-website.vercel.app

Everything you made, in one vault.

Personify turns the exports already sitting on your laptop—chats, mail, code, notes, and files—into a local Postgres archive you can search, connect, and hand to your agents without handing away ownership.

Research previewRead the install guideExplore the vault

Local Postgres · Nine source adapters · Read-only MCP · Your hardware

Flat isometric illustration of white and coral archive cartridges entering a secure local vault, with a knowledge graph and agent port emerging on the other side

Fig.01 — The local archive

Raw copied · never moved

Bring the archive

ChatGPT
Claude
Gmail
Discord
Notion
GitHub
X
Files
Takeout

Export files in · One normalized item schema out

Vault overview

One place to see what your archive actually contains.

The real Personify dashboard leads with counts, source distribution, and ingestion history. Every bar opens the records behind it; every run keeps its parser version and outcome.

personify / personal/Dashboard
Search the vault

Vault overview

Dashboard

Personal data ingested from your services.

Items

5,796

Exports

5

Sources

4

Accounts

5

Runs

5

Items per source
Claude
2,906
X (Twitter)
2,567
GitHub
321
Files
2

Immutable intake

Import the source. Keep the evidence.

Register an export once. Personify copies and hashes the archive, selects the source parser, writes a normalized item stream, then optionally builds vectors and graph entities. Replace runs rebuild derived data—not history.

sourceWhere the record came from
accountWhich identity owned it
native_idIts source-system identity
content_hashStable fallback deduplication
metadataSource-specific context, intact

Register an export

raw → normalized → queryable

Local only

Source parser

ChatGPT
Claude
GitHub
Files
Archive pathC:\Exports\chatgpt-data.zip
Accountpersonal
✓ ingest✓ embeddings✓ graph

Pipeline complete

One immutable source, four reversible stages.

18.4s

Register

archive copied · SHA-256

ok

Ingest

2,906 items normalized

ok

Embed

384-dim local vectors

ok

Graph

1,248 entities extracted

ok

seen

2,927

inserted

2,906

skipped

21

Search the whole you

Find the thing you remember, not just the words.

Switch between literal full-text rank and local vector similarity, narrow by source, then open a result down to its account, native ID, ingestion run, hash, body, tags, and metadata.

Find

Search

that decision about pgvector
3 results · similarityall sources

Context graph

Follow the connection back to proof.

Personify extracts projects, people, tools, decisions, files, topics, and more. The graph is not a decorative map: each relationship carries source evidence, aliases, and suggested follow-up queries.

22

entity types

1

hop at a time

100%

evidence linked

Knowledge graph

Walk the neighborhood.

Search entities by name or alias…
Projectpersonal vault

Entity

personal vault

Project · 12 aliases

Evidence

“Keep retrieval inside the same local ownership boundary.”

Neighborhood

7 related · 11 edges

depth 1 · evidence linked

Agent access

Give agents context, not custody.

The same search, timeline, item, and graph surfaces ship as a read-only MCP server over stdio, plus a UI-gated local HTTP transport. Claude Desktop, Codex, and local clients can ask; the vault remains yours.

Agent access

read-only MCP · active vault: personal

running

HTTP transport

Local endpoint, explicit switch.

endpoint
http://127.0.0.1:18765/mcp
uptime
02:14:36
requests
184
sessions
2
stdio entry point
$ vault --vault personal mcp

Tool surface

13 tools · 6 resources

read only

search

full-text retrieval

semantic_search

meaning across sources

timeline

items by time window

get_item

source-linked record

entity_neighborhood

graph traversal

+ 8 more tools

Recent requests

semantic_search218ms12 results
entity_context84ms7 related
get_item31msitem 4821

Ownership boundary

01

Raw stays raw

Exports are copied into immutable storage and SHA-256 verified. The source archive is never rewritten.

02

Identity survives

Native IDs deduplicate first; content hashes cover the rest. Re-importing the same export does not duplicate your history.

03

Enrichment is optional

Parsing is standard. Embeddings and graph extraction are opt-in stages you can re-run without touching the original.

04

Vaults stay separate

Named vaults get their own Postgres database and filesystem root, so work and personal context never have to mix.

Open the vault

Start with one export

Your archive should answer to you.

Set up the local vault, register the exports you already own, and give your agents a context layer that stays on your hardware.

Install PersonifyReview the pipeline

FastAPI · Postgres 17 · pgvector · Local filesystem · MCP