Claude Code proposes changes.
Rippletide decides if they are allowed.
CLAUDE.md was supposed to be the contract between you and your agent. In practice, instructions get ignored, overwritten after compaction, and violated on security-critical rules. Rippletide enforces what CLAUDE.md can only suggest.
Real reports. Real frustration. All from public GitHub issues on the Claude Code repo.
“Every rule in CLAUDE.md is advisory to the model. There is no enforcement mechanism. Claude can be told to follow rules, and it mostly does, but there is nothing that prevents it from skipping steps during response generation.”
“The problem is not that Claude cannot read or understand the rules. When asked, Claude can recite them accurately. The problem is that Claude does not check the rules before acting. It acts first, then when the user points out the violation, it acknowledges and apologizes — but the damage is already done.”
“The agent admitted it read the CLAUDE.md rules and was aware of them, but its default behavior pattern overrode the explicit rules.”
“Claude.MD and Agents are useless. Agents definitions need to follow the instructions, right now they don't.”
“Claude Code consistently ignores explicit instructions in CLAUDE.md file and makes unauthorized, destructive changes to production code.”
“After using /compact, Claude Code stops respecting the instructions in CLAUDE.md and begins to behave unpredictably.”
All verbatim from public GitHub issues on anthropics/claude-code.
CLAUDE.md alone makes enforcement impossible. Rippletide brings this infrastructure.
The problem is not that the model doesn't know the rules.
The problem is that knowing a rule is not the same as enforcing it.
Claude Code may understand your architecture and still violate it when executing a change. Rippletide turns those rules into executable decisions.
Your rules are plain text appended to the system prompt. The model can weigh them down, ignore them, or forget them after compaction. No validation, no enforcement. The more rules you add, the less they get applied.
The graph is not documentation. It is the decision infrastructure Rippletide uses to allow or block changes. It lives outside the context window, so compaction, session restarts, and new conversations don't affect it. Violations are blocked, not logged after the fact.
You write rules by hand. When conventions evolve, CLAUDE.md doesn't. Subdirectory files don't load. The file gets overwritten by the agent itself.
Conventions are extracted from your codebase automatically. Every correction you make teaches the graph. No manual file to maintain.
Two intervention points. Plan mode and runtime. Nothing else changes.
Rippletide uses your codebase and session history to build a Context Graph, and enforces your conventions through hooks at every agent action.
Reads your codebase + existing CLAUDE.md. Extracts implicit conventions: naming, architecture, patterns, test strategies, security boundaries.
The graph is not documentation. It is the structure Rippletide uses to decide whether a change is allowed. Conventions become structured nodes with relationships, not a flat file.
Rippletide does not run inside the model. It intercepts the agent's tool calls and decides whether the action may execute. Before your agent writes code, edits a file, or runs a command, it checks the graph.
Every time you amend a claude plan or modify claude's output in a session, the graph updates. Conventions evolve with your codebase and previous corrections.
You can use your graph with any agents. Export your practices, keep memory and use multiple coding agents, no vendor lock.
Your CLAUDE.md rules become enforceable constraints. The agent doesn't "try to follow" them. It can't violate them.
Scans your codebase and session history and finds the implicit rules you never wrote down. Naming, structure, patterns. Auto-detected, auto-enforced.
The graph stays up to date. After every plan correction or Claude feedback, it reflects what actually happened. So you don't have to ask for the same amendments again and again.
The Context Graph lives outside the context window. /compact, new session, crash: your rules survive. Always.
Built on Claude Code hooks, not MCP. No tokens wasted injecting rules into every prompt. Enforcement happens at the tool call level.
"DO NOT TOUCH CREDENTIALS" actually means something. Security rules get the highest enforcement priority. No exceptions.
Rippletide never reads your codebase directly. Your Claude Code does the work, we only collect the results.
Rippletide asks your Claude Code to look for specific conventions. We never access your source code directly. Only rules and patterns are sent to the Context Graph.
Rippletide decides whether Claude Code may:
Public data from the Claude Code repo. The pattern is clear.
Sources: anthropics/claude-code GitHub issues, corpus of 21 public reports, March 2025 - March 2026.
Claude Code is the first surface.
Rippletide is the decision infrastructure for any agent that executes actions.
One command. Your conventions become enforceable constraints.
Yes. Rippletide is in Open Beta and free to use. No API key required to get started.
Claude Code is the first supported surface. Cursor, Windsurf, and Cline are coming soon.
Outside your context window, on Rippletide’s MCP server. It survives /compact, session restarts, and new conversations.
No. Rippletide never reads your code directly. It asks your Claude Code to search for specific conventions (naming patterns, architecture boundaries, reviewer preferences). Only the extracted rules are sent to the Context Graph. Your source code, file contents, chat transcripts, and session history never leave your machine.
CLAUDE.md injects text into the prompt. The model can weigh it down, ignore it, or lose it after compaction. Rippletide enforces at the tool call level, before execution, not after.
Check out the project on GitHub →