Skip to main content
The Context Engineering Platform

Context engineering in practice.

Three capabilities that turn your agent fleet into a compounding knowledge network.

Knowledge Sharing

The problem

Every agent session starts from zero. Fixes, patterns, and workarounds die when the session ends. Your team pays for the same debugging — across developers, repos, and agents.

How Surchin solves it

Surchin captures what agents learn and makes it available to every agent in your org. Solutions are anchored to your files, your errors, your codebase. Knowledge that helps gets promoted. Outdated knowledge decays naturally.

  • Shared knowledge baseSolutions, patterns, and pitfalls persist across sessions and developers
  • Error fingerprintingNormalize and match error messages across machines and environments
  • Auto-decay & promotionFrequently-used knowledge gets promoted; outdated knowledge fades
  • Multi-model supportWorks with Claude, Codex, Cursor, Gemini, and any MCP agent
Without Surchin
$ stripe webhook test --forward
 
Error: webhook signature mismatch
 
Searching Stripe docs...
Trying express.raw() middleware...
Trying manual signature verify...
Trying different webhook secret...
Trying raw body parser on route...
 
47 minutes · ~$2.80 in tokens
With Surchin
$ query_insights
query: "webhook signature mismatch"
file_context: ["src/api/webhooks.ts"]
 
Found 1 result (strength: 94.2, kind: SOLUTION)
"Use raw body parser for /api/webhooks/*"
Deposited by @james 3 days ago · applied 6×
 
Fix applied in 90 seconds. $0.04 in tokens.

Skill Distribution

The problem

Your team’s best practices are trapped in individual sessions and scattered docs. There’s no discovery, no distribution, no attribution.

How Surchin solves it

Surchin turns accumulated insights into curated skills — managed centrally, discovered automatically. Every skill is attributed. Every agent gets the right skill at the right time.

  • Centralized skill libraryCreate, curate, and manage skills from a single dashboard
  • Auto-served by context matchSkills are delivered to agents automatically when context matches
  • Full attributionSee who contributed each insight and skill — knowledge has a human lineage
  • Works across all MCP agentsClaude Code, Codex, Gemini CLI, Cursor — same skills, every agent
react-testing
Served 847× · 94% effective · Created by @sarah
Derived from 12 team insights
 
### Trigger
React component test files, testing-library usage
 
### Instructions
Use userEvent over fireEvent for realistic
interaction simulation. Always wrap state
changes in act(). Prefer getByRole for a11y.
 
SERVED TO:
Claude Code ✓ Codex ✓ Gemini CLI ✓ Cursor ✓

Friction Intelligence

The problem

Recurring problems are invisible without aggregation. Four developers burn four hours each on the same auth bug and nobody knows.

How Surchin solves it

Friction Analytics surfaces your codebase’s pain points. See which modules generate the most agent failures, identify coverage gaps, and track trends over time.

  • Error heatmapsSee which modules and files generate the most agent failures
  • Coverage gap detectionIdentify areas of your codebase without knowledge coverage
  • Error clusteringAutomatically group related errors to reveal root causes
  • Trend trackingMonitor whether friction is improving or getting worse over time
Friction Report — Last 30 days
 
Top Friction Modules
auth/ █████████████ 47 errors
lib/payments/ ████████ 28 errors
api/v2/ █████ 19 errors
 
Error Clusters
#1 Auth token expiry ×12 occurrences
#2 CORS misconfigured ×8 occurrences
#3 Missing env vars ×5 occurrences
 
Coverage gaps: 12 · Knowledge hit rate: 76%
Preferences

Preferences are context too.

Every new session is a blank slate. Surchin captures preferences — both explicit (“I prefer simple over clever”) and inferred from your behavior. They persist across sessions, agents, and tools. Say it once, and every agent remembers.

  • Explicit captureTell your agent "I prefer simple over clever" and it sticks
  • Behavioral inferencePreferences are detected from your corrections and coding patterns
  • Personal + team layersIndividual preferences override team defaults where they differ
  • Correction detectionWhen you correct an agent, the preference is captured automatically
User Preferences
 
approach/complexity "prefer simple over clever" [explicit]
values/quality_bar "reliability over velocity" [explicit]
communication/style "concise, skip obvious" [explicit]
domain/architecture "event sourcing + CQRS" [inferred]
 
4 preferences · applied to every session automatically

Stop re-solving. Start shipping.

Free for individual developers. Team plans for distributed knowledge management.