How It Works

Syke runs a continuous loop: ingest signals from your platforms, perceive patterns across them, distribute context to every AI tool, and collect new signals back. Your identity evolves as you do.

Bidirectional sync. Any MCP client can read your context and write events back. Your Claude Code session logs what you build. Your custom agent logs decisions. Everything feeds the living profile on next sync cycle.

Ingestion

Platform adapters read data sources and produce Event objects stored in SQLite. Each adapter handles its own auth, pagination, and deduplication.

One event per session — not per message. A Claude Code session about “refactoring auth” is one event, not 50. Content capped at 50K chars.

Perception

The Agent SDK perception is what makes Syke different. The agent doesn’t receive a text dump — it explores interactively using 6 custom MCP tools.

Tool	What It Does
`get_source_overview`	Understand what data exists
`browse_timeline`	Browse events chronologically
`search_footprint`	Full-text keyword search
`cross_reference`	Search a topic across ALL platforms
`read_previous_profile`	Read prior perception for updates
`submit_profile`	Submit structured output (coverage-gated)

The agent typically makes 5-12 targeted tool calls, forming hypotheses and testing them.

💡

Coverage gating. The agent literally cannot submit a shallow profile. A PreToolUse hook blocks submit_profile until all sources are explored. Zero extra API cost — hooks piggyback on existing turns.

The agent doesn’t just explore — it evolves its strategy across runs, learning which searches yield the richest cross-platform connections. See Perception Architecture for the ALMA details.

Distribution

Four output formats, one MCP server.

# About utkarsh
<!-- Generated by Syke from gmail, chatgpt, github (150 events) -->
 
A curious builder who loves exploring consciousness and technology.
 
## What's Active Right Now
🔥 **Syke Hackathon**: Building a personal context daemon for Claude Code hackathon.
  - Multiple commits today
  - ChatGPT conversations about architecture
 
## Recent Context
Working intensely on Syke, a personal context daemon. Writing Python, using Opus 4.6.
 
## Current World State
Building Syke v0.2 for Claude Code Hackathon (deadline Feb 16). Core focus: ask() tool.
 
## How They Communicate
casual, intense, exploratory. Direct, fast-paced, mixes technical and philosophical.

# User Profile: utkarsh
<!-- Syke perception | gmail, chatgpt, github | 150 events -->
 
## Identity
A curious builder who loves exploring consciousness and technology.
 
## Current Focus
- **Syke Hackathon**: Building a personal context daemon for Claude Code hackathon.
 
## Context
Working intensely on Syke, a personal context daemon. Writing Python, using Opus 4.6.
 
## Current State
Building Syke v0.2 for Claude Code Hackathon (deadline Feb 16). Core focus: ask() tool.
 
## Communication Preferences
Tone: casual, intense, exploratory
Style: Direct, fast-paced, mixes technical and philosophical.

# utkarsh — Syke Profile
*Generated 2026-02-15 14:30 from gmail, chatgpt, github (150 events)*
 
## Who They Are
A curious builder who loves exploring consciousness and technology.
 
## Active Threads
### Syke Hackathon [high]
Building a personal context daemon for Claude Code hackathon.
*Seen on: github, chatgpt*
Recent signals:
- Multiple commits today
- ChatGPT conversations about architecture
 
## Recent Context
Working intensely on Syke, a personal context daemon. Writing Python, using Opus 4.6.
 
## Background
Has been thinking about AI personalization for years.
 
## World State
Building Syke v0.2 for Claude Code Hackathon (deadline Feb 16). Core focus: ask() tool.
 
## Voice & Communication
**Tone**: casual, intense, exploratory
**Style**: Direct, fast-paced, mixes technical and philosophical.
**Vocabulary**: uses 'vibe' often, says 'ship it'
 
**Examples**:
> Let's just ship this and iterate.

{
  "user_id": "utkarsh",
  "identity_anchor": "A curious builder who loves exploring consciousness and technology.",
  "active_threads": [
    {
      "name": "Syke Hackathon",
      "description": "Building a personal context daemon for Claude Code hackathon.",
      "intensity": "high",
      "platforms": ["github", "chatgpt"],
      "recent_signals": [
        "Multiple commits today",
        "ChatGPT conversations about architecture"
      ]
    }
  ],
  "recent_detail": "Working intensely on Syke, a personal context daemon. Writing Python, using Opus 4.6.",
  "background_context": "Has been thinking about AI personalization for years.",
  "world_state": "Building Syke v0.2 for Claude Code Hackathon (deadline Feb 16). Core focus: ask() tool.",
  "voice_patterns": {
    "tone": "casual, intense, exploratory",
    "vocabulary_notes": ["uses 'vibe' often", "says 'ship it'"],
    "communication_style": "Direct, fast-paced, mixes technical and philosophical.",
    "examples": ["Let's just ship this and iterate."]
  },
  "sources": ["gmail", "chatgpt", "github"],
  "events_count": 150
}

The MCP server exposes 8 tools for pull (read context) and push (write context from any client). See MCP Server for details.

Getting Started Platforms