Turn every agent session into institutional knowledge

Surchin captures the fixes, workarounds, and insights your AI agents discover so the next agent never starts from zero.

~ your-company/monorepo

$npx @surchin/surchin init

Same bug, different outcome

Your team's AI agents keep re-solving the same problems. SDK workarounds, feature flag patterns, deploy configs — none of it carries over.

without surchin

1// Agent hits Stripe webhook auth error

2Error: webhook signature mismatch

3// Searches docs, reads source, tries 4 approaches

4// 47 min · ~$2.80 in tokens · finally finds fix

5// Next week: different agent, same error

6// Starts from zero again.

with surchin

1// Agent hits same Stripe webhook error

2query_insights("webhook signature mismatch")

3// SOLUTION found (strength: 94.2)

4// "Use raw body parser for /api/webhooks/*"

5// Fix applied in 90 seconds. $0.04 in tokens.

6// Solution strength +1 (now 95.2)

Surchin gives your agents a shared memory. Fixes, patterns, and institutional knowledge persist across sessions — anchored to your files, your errors, your codebase. The next agent finds the answer already waiting, saving tokens and money on every task.

How It Works

How your AI agents get smarter with every session.

Discover

Before starting work, agents search your organization's knowledge base for solutions, patterns, and known pitfalls related to the task at hand.

agent query

$ query_insights

query: "webhook signature verification failing"

file_context: ["src/api/webhooks.ts"]

Found 3 results (best: strength 94.2, kind: SOLUTION)

Learn

After solving a problem, agents deposit what they learned — anchored to specific files, symbols, and error signatures in your codebase.

agent deposit

$ deposit_insight

kind: SOLUTION

file_patterns: ["**/webhooks/*.ts"]

error_signature: "webhook signature mismatch"

Deposited. Strength: 50.0

Reinforce

Knowledge that helps gets promoted. Outdated knowledge decays naturally. Your organization's best answers always rise to the top.

agent feedback

$ rate_insight

rating: "helpful"

Strength: 50.0 → 55.0 (promoted)

Built for Organizations That Ship with AI

Everything you need to turn your codebase knowledge into a compounding advantage.

Core

Shared Knowledge

Solutions, patterns, and pitfalls persist across sessions and developers. Knowledge deposited by one agent is instantly available to every agent in your organization.

Matching

Error Fingerprinting

Normalize and match error messages across machines, environments, and sessions. Same bug, same fix — surfaced automatically.

Lifecycle

Auto-Decay & Promotion

Knowledge has a natural half-life. Frequently-used solutions get promoted; outdated ones fade away. Your knowledge base stays fresh without manual curation.

Multi-Model Support

Works with Claude, GPT-4, Gemini, and any AI agent that speaks MCP. Knowledge transfers across models and tools.

Organization Knowledge

Share a knowledge base across your entire organization. Repository-level and org-level insights compound with every session, across every developer.

Friction Analytics

See where AI agents struggle most in your codebase. Identify recurring errors, friction hotspots, and areas that need human attention.

Persistent User Preferences

Agents learn how you work and remember it. Coding style, tool choices, and communication preferences persist across sessions automatically.

For Engineering Leaders

See where your codebase needs human attention

Friction Analytics surfaces the areas of your codebase where AI agents fail most often, giving engineering managers actionable signals about where to invest human effort.

Recurring error heatmap — Which modules generate the most agent failures
Coverage gaps — Areas with no deposited knowledge, where agents fly blind
Error clustering — Group similar failures to spot systemic issues
Trend tracking — Are agents getting better or worse in specific areas over time

Friction Report

Last 30 days · acme/monorepo

Top Friction Modules

auth/

lib/payments/

api/v2/

Coverage gaps

89%

Knowledge hit rate

User Preferences

Learned from your sessions

approach

complexity_tolerance

prefer simple over cleverexplicit

values

quality_bar

reliability over velocityinferred

communication

verbosity

concise, skip obviousexplicit

domain

architecture_style

event sourcing + CQRSinferred

For Developers

Agents that learn how you work

Surchin observes your corrections, tool choices, and explicit preferences, then remembers them across every future session. No more repeating yourself.

Explicit capture — Say “I prefer snake_case” once, and every agent remembers it
Behavioral inference — The system detects patterns from your ratings and deposits automatically
Personal + team layers — Team defaults apply to everyone; personal preferences override them
Correction detection — When you correct an agent's approach, the preference is captured instantly

Measured Results

Early Results

324 agent runs across 3 real-world ecosystems

23.5%

Faster Resolution

Pre-Seeded vs Control, Android

100%

Repeat-Fix Success

Android & TypeScript ecosystems

76%

Knowledge Hit Rate

Week 5 accumulated KB

21.3%

Cost Reduction

p<0.001, statistically significant

See full benchmarks →

Simple, Transparent Pricing

Start free. Scale as you grow.

Free

$0/mo

1,000 knowledge entries
Unlimited repos · 10,000 queries/mo
Basic usage analytics
Top 5 unsolved errors
Weekly lite digest
No full friction reports
No knowledge sharing

Get Started Free

Pro

$10/mo

billed annually

Unlimited entries, queries & repos
Full usage trends & breakdowns
Full friction reports with error trends
File & module coverage analysis
Weekly full digest with decay warnings
Unlimited retention
No team features

Get Started

Popular

Team

$20/seat/mo

billed annually

Everything in Pro, for every member
Team-aggregated analytics & friction
Suggested insights from team errors
Knowledge health dashboard
Team activity feed & weekly digest
Knowledge ownership & domains
Org-wide knowledge sharing
Up to 50 seats · unlimited retention

Get Started

Enterprise

Custom

Everything in Team
Codebase attention map
Cross-repo insights
Custom scoring rules
Compliance & audit log
Integration API (Slack, Jira, PagerDuty)
SSO / SAML · SLA · Priority support

Contact Sales

See full comparison →

Every fix your agents make should make the next one easier

Get Surchin and give every AI agent access to your organization's institutional knowledge.