Protocols - How lean-ctx Communicates

lean-ctx defines 5 protocols that govern how context is managed, compressed, communicated, and shared between agents. They activate automatically based on session state, but understanding them helps you get the most out of the system.

Protocol Map

Protocol	Full Name	When It Activates	What It Does
`CEP`	Context Engineering Protocol	Every session (system prompt)	5 rules for efficient context usage: act first, delta only, structured notation, one line per action, quality anchor
`CCP`	Context Checkpoint Protocol	Every ~15 tool calls (auto) or manual	Creates compressed session snapshots that survive context window truncation
`TDD/CRP`	Terse Data Density / Context Reduction Protocol	When `CRP_MODE=tdd` is set	Maximum compression: abbreviated keywords, symbol maps, diff-only output
`CLP`	Context Layered Preload	Session start with `ctx_preload`	Intelligently pre-loads relevant files based on task description
`A2A`	Agent-to-Agent	Multi-agent sessions	Message bus, shared cache, task delegation, diary persistence

CEP - Context Engineering Protocol

CEP is injected into every session's system prompt. It establishes 5 core rules that reduce token waste from the LLM itself:

ACT FIRST: Execute tool calls before explaining. Avoid narration before action.
DELTA ONLY: Use file references (F1, F2...) instead of repeating paths. Show only changed code.
STRUCTURED: Use + (add), - (remove), ~ (modify) notation.
ONE LINE PER ACTION: Keep each action to a single line for scan-ability.
QUALITY ANCHOR: Never skip edge cases, never use mock data, never placeholder.

Full CEP documentation →

CCP - Context Checkpoint Protocol

CCP automatically creates compressed session snapshots every ~15 tool calls. These checkpoints capture:

Cached files with their F-references and content hashes
Active session state (task, workflow, knowledge)
File signatures for quick recovery

When the context window is truncated (common in long sessions), the checkpoint allows the agent to recover its full working state without re-reading all files.

Full CCP documentation →

TDD / CRP - Terse Data Density Mode

Activated when CRP_MODE=tdd is set in the environment or config. Applies maximum compression to all lean-ctx output:

Abbreviated keywords: fn, cfg, impl, deps, req, res, ctx, err
Type shortcuts: :s (string), :n (number), :b (boolean)
Symbol map tables (§MAP) for long identifiers → short IDs (α1, α2...)
Diff-only output (+/-/~ notation)
Budget target: ≤150 tokens per response, zero narration

Full TDD documentation →

CLP - Context Layered Preload

CLP activates when you call ctx_preload at session start. It analyzes your task description and the project structure to pre-load the most relevant files in optimal read modes.

Layers

Layer 1 (Entry): Files directly mentioned or most relevant to the task - read in full mode
Layer 2 (Context): Dependencies and imports of Layer 1 files - read in map or signatures mode
Layer 3 (Reference): Transitive dependencies - registered as reference mode (no content loaded)

ctx_preload task="fix the auth refresh token bug"
→ Preloaded 7 files in 3 layers:
  L1 (full):    src/auth/service.rs, src/auth/token.rs
  L2 (map):     src/http/middleware.rs, src/db/sessions.rs, src/config.rs
  L3 (ref):     src/http/server.rs, src/main.rs
  Total: ~2400 tokens (vs ~12000 tokens for full reads)

A2A - Agent-to-Agent Protocol

A2A enables coordination between multiple AI agents working on the same project. It provides an agent registry, message bus, task delegation, shared cache, and persistent diaries — everything needed for multi-agent workflows on a single codebase.

Full multi-agent documentation →