Architecture memory for AI coding agents
Chronicle Pro helps enterprise engineering teams make AI coding agents reliable in large codebases by giving them self-hosted, Git-versioned architecture memory.
Currently onboarding a small number of enterprise design partners
Built for MCP-compatible AI tools, including Claude Code and Cursor
AI coding agents are entering enterprise codebases — but they still lack architecture context.
Ask AI to inspect files manually
Missing cross-repo context
Hallucinated dependencies
Stale diagrams
Senior engineers become bottlenecks
❯ what depends on BattleEvent?
reading arena.service.ts...
reading arena.module.ts...
reading battle-result.producer.ts...
grep BattleEvent across 41 files...
...8 file reads later...
"BattleEvent is used in ArenaService and
possibly the spectators service"
⚠ missed: Kafka consumer, queue processor
Persistent architecture memory
Source-backed dependency graph
Impact paths across APIs/events/services
Diagrams generated from current knowledge
AI agents answer with project context
❯ what depends on BattleEvent?
■ 3 services, 5 edges, 2 layers
→ ArenaService.create() — writes model
→ BattleResultProducer → kafka:battle-results
→ BattleResultConsumer ← kafka:battle-results
→ SpectatorService.recordBattle()
✓ 0.02s — 0 file reads — 5 evidence assertions
Services, APIs, events, models, flows, ownership — extracted from code and connected.
Answers grounded in extracted facts and source evidence, not temporary chat memory.
Tied to Git revisions. Knowledge updates when code changes. Full history and rollback.
Works with Claude, Cursor, and any MCP-compatible AI tool.
The open-source version helps an agent understand a project.
The Pro version gives organizations shared, governed and federated architecture memory.
Each product or team keeps its own architecture memory, ownership model and Git history.
Remote agents connect via the MCP protocol. No new APIs to learn — Chronicle talks to Chronicle natively.
See what each team declares in their manifest versus what the code graph actually proves. Spot architecture drift before it becomes a production incident.
One dashboard view of every architecture source — local repos, remote agents, connection status, scan freshness, and unresolved cross-boundary references.
"What depends on ArenaService?"
Chronicle builds a knowledge graph from your code — 100 nodes, 119 edges
"What breaks if BattleEvent model changes?"
❯ what breaks if I change the BattleEvent model?
...
■ Impact Analysis — 3 services, 7 edges, 2 layers
Affected:
ArenaService.create() → writes BattleEvent
ArenaService.getHistory() → reads BattleEvent
BattleResultProducer → publishes to kafka:battle-results
BattleResultConsumer ← consumes from kafka:battle-results
SpectatorService.recordBattle() → processes event
Evidence:
arena-api/src/arena/arena.service.ts
arena-api/src/arena/battle-result.producer.ts
spectators-api/src/spectators/battle-result.consumer.ts
✓ answered in 0.02s — 0 file reads — 5 evidence assertions
AI agents query the graph for source-backed answers — not file reads
"Trace the flow from POST /arena/attack to spectators"
Trace request flows across service boundaries — HTTP, Kafka, WebSocket
Navigate a real 4-service architecture graph — C1 system view, C2 service dependencies, C3 components
Currently onboarding a small number of enterprise design partners.
Request design partner access →Local project memory
Single repo / project
One knowledge base
Local graph queries
—
—
Local developer workflow
Community support
View on GitHubShared enterprise memory
Multi-repo, multi-team
Federated team knowledge bases
Cross-agent federated queries via MCP
Declaration vs Reality drift detection
Sources dashboard with remote agent status
Platform / architecture team workflow
Enterprise onboarding & support
Request accessCompare what teams declare in manifests against what the code graph actually proves. Catch governance gaps automatically.
Deploy inside your infrastructure. Your code and graph stay under your control.
Architecture state tied to Git revisions. Full history and rollback.
Remote Chronicle agents connect as MCP servers. Read-only queries with trust policies. No new protocol to learn.
Designed for private deployment. External model usage configured by your security policy.
Traceable knowledge changes and evidence. Every graph fact backed by source file, line, and assertion.
Built for platform teams, staff engineers and enterprise architects managing complex multi-repo systems.
Teams building internal developer platforms and shared tooling.
Architects maintaining system-wide dependency knowledge and governance.
Leaders driving safe AI adoption across engineering orgs.
We are selecting a small number of enterprise design partners to shape Chronicle Pro for real multi-repo engineering environments.
8-week pilot · 1–3 repositories · Self-hosted deployment
Fixed architecture questions · Weekly sync · Clear success criteria
Start with one product slice, then expand to federated team knowledge bases after validation.
Selected teams will receive a scoped pilot proposal after an initial discovery call.