Your AI doesn't know you

Every conversation starts cold. Syke synthesizes your digital footprint into psyche-level context — one perception, every AI agent knows you.

Get StartedTechnical Details

The difference context makes

Same model. Same question. The only difference is Syke.

Cold start
I'd be happy to help you with your project! Could you tell me more about what you're working on? What programming language are you using? What's the goal of the project? Any specific requirements or constraints I should know about?
Syke-enhanced
Based on your recent coding sessions — you're building a CLI tool in Python with Click. Your architecture discussions show you're considering an agent-based approach with MCP integration, and your commit history confirms you've been iterating on the perception pipeline. Here's what I'd suggest for the next step...
coding sessions
conversations
commit history

Your digital footprint, unified

Syke ingests from the platforms where you think, code, and communicate.

✓

Claude Code

Sessions, debugging, architecture decisions

✓

Claude Desktop Cowork

Extended conversations, research, strategy

✓

Claude Web

Browser-based Claude conversations

✓

GitHub

Repos, PRs, commits, issues

✓

Gmail

Email threads and correspondence

coming soon

Other AI Platforms

Any platform via export

✓

Other Chat Exports

ZIP/JSON exports from any tool

Built for developers

Simple, private, and extensible. Context that works the way you do.

Privacy First

All data stays local in SQLite on your machine. Nothing leaves unless you run perception.

Open Source

MIT licensed. Inspect every line. No telemetry, no tracking, no surprises.

MCP Native

8 tools, works with any MCP client. Claude Code, Claude Desktop, or your own agent.

One Perception, Every Agent

Run Syke once. Your identity follows you to every AI tool, every conversation.

How it fits together

Three steps from scattered conversations to unified identity

Step 1

Ingest

Reads your digital footprint from multiple platforms

Claude Code, Claude Desktop, GitHub, Gmail, chat exports

4 sources
Step 2

Perceive

Opus 4.6 explores your footprint with 6 MCP tools

Agent SDK + extended thinking, coverage-gated quality

Agent SDK
Step 3

Distribute

Your identity goes wherever your AI agents are

MCP server, CLAUDE.md injection, JSON/Markdown export

8 MCP tools
MCP Server — 8 tools
ask(question)Agentic natural language queries
get_profile(format)Identity in 4 formats
query_timeline(...)Events by date/source
search_events(query)Keyword search
get_event(id)Full event content
push_event(...)Federated push from any client
push_events(json)Batch push
get_manifest()Data statistics
Example
$ uvx syke setup --yes

Detecting sources...
  Claude Code: 47 sessions found
  GitHub: <username> (142 repos)
  Gmail: authorized via gog

Ingesting... 2,847 events
Perceiving... Opus 4.6 + Agent SDK
  Coverage: 100% (3/3 sources)

Done. Profile written to ~/.syke/

Get started

One command to perceive. Works with Claude Code, Claude Desktop, any MCP client.

$ uvx syke setup --yes
Or tell your agent
Study this and set it up for me: github.com/saxenauts/syke — run uvx syke setup --yes

Paste into Claude Code, Cursor, or any agent with terminal access.

→ DocsGitHubPyPI