CASS - Coding Agent Session Search

Unified, high-performance CLI/TUI to index and search your local coding agent history. Aggregates sessions from 11 agents: Codex, Claude Code, Gemini CLI, Cline, OpenCode, Amp, Cursor, ChatGPT, Aider, Pi-Agent, and Factory (Droid).

CRITICAL: Robot Mode Required for AI Agents

NEVER run bare cass - it launches an interactive TUI that blocks your session!

# WRONG - blocks terminal
cass

# CORRECT - JSON output for agents
cass search "query" --robot
cass search "query" --json  # alias

Always use --robot or --json flags for machine-readable output.

---

Quick Reference for AI Agents

Pre-Flight Check

# Health check (exit 0=healthy, 1=unhealthy, <50ms)
cass health

# If unhealthy, rebuild index
cass index --full

Essential Commands

# Search with JSON output
cass search "authentication error" --robot --limit 5

# Search with metadata (elapsed_ms, cache stats, freshness)
cass search "error" --robot --robot-meta

# Minimal payload (path, line, agent only)
cass search "bug" --robot --fields minimal

# View source at specific line
cass view /path/to/session.jsonl -n 42 --json

# Expand context around a line
cass expand /path/to/session.jsonl -n 42 -C 5 --json

# Capabilities discovery
cass capabilities --json

# Full API schema
cass introspect --json

# LLM-optimized documentation
cass robot-docs guide
cass robot-docs commands
cass robot-docs schemas
cass robot-docs examples
cass robot-docs exit-codes

---

Why Use CASS

Cross-Agent Knowledge Transfer

Your coding agents create scattered knowledge:

• Claude Code sessions in ~/.claude/projects
• Codex sessions in ~/.codex/sessions
• Cursor state in SQLite databases
• Aider history in markdown files

CASS unifies all of this into a single searchable index. When you're stuck on a problem, search across ALL your past agent sessions to find relevant solutions.

Use Cases

# "I solved this before..."
cass search "TypeError: Cannot read property" --robot --days 30

# Cross-agent learning (what has ANY agent said about X?)
cass search "authentication" --robot --workspace /path/to/project

# Agent-to-agent handoff
cass search "database migration" --robot --fields summary

# Daily review
cass timeline --today --json

---

Command Reference

Indexing

# Full rebuild of DB and search index
cass index --full

# Incremental update (since last scan)
cass index

# Watch mode: auto-reindex on file changes
cass index --watch

# Force rebuild even if schema unchanged
cass index --full --force-rebuild

# Safe retries with idempotency key (24h TTL)
cass index --full --idempotency-key "build-$(date +%Y%m%d)"

# JSON output with stats
cass index --full --json

Search

# Basic search (JSON output required for agents!)
cass search "query" --robot

# With filters
cass search "error" --robot --agent claude --days 7
cass search "bug" --robot --workspace /path/to/project
cass search "panic" --robot --today

# Time filters
cass search "auth" --robot --since 2024-01-01 --until 2024-01-31
cass search "test" --robot --yesterday
cass search "fix" --robot --week

# Wildcards
cass search "auth*" --robot          # prefix: authentication, authorize
cass search "*tion" --robot          # suffix: authentication, exception
cass search "*config*" --robot       # substring: misconfigured

# Token budget management (critical for LLMs!)
cass search "error" --robot --fields minimal              # path, line, agent only
cass search "error" --robot --fields summary              # adds title, score
cass search "error" --robot --max-content-length 500      # truncate fields
cass search "error" --robot --max-tokens 2000             # soft budget (~4 chars/token)
cass search "error" --robot --limit 5                     # cap results

# Pagination (cursor-based)
cass search "TODO" --robot --robot-meta --limit 20
# Use _meta.next_cursor from response:
cass search "TODO" --robot --robot-meta --limit 20 --cursor "eyJ..."

# Match highlighting
cass search "authentication error" --robot --highlight

# Query analysis/debugging
cass search "auth*" --robot --explain    # parsed query, cost estimates
cass search "auth error" --robot --dry-run  # validate without executing

# Aggregations (server-side counts)
cass search "error" --robot --aggregate agent,workspace,date

# Request correlation
cass search "bug" --robot --request-id "req-12345"

# Source filtering (for multi-machine setups)
cass search "auth" --robot --source laptop
cass search "error" --robot --source remote

# Traceability (for debugging agent pipelines)
cass search "error" --robot --trace-file /tmp/cass-trace.json

Session Analysis

# Export conversation to markdown/HTML/JSON
cass export /path/to/session.jsonl --format markdown -o conversation.md
cass export /path/to/session.jsonl --format html -o conversation.html
cass export /path/to/session.jsonl --format json --include-tools

# Expand context around a line (from search result)
cass expand /path/to/session.jsonl -n 42 -C 5 --json
# Shows 5 messages before and after line 42

# View source at line
cass view /path/to/session.jsonl -n 42 --json

# Activity timeline
cass timeline --today --json --group-by hour
cass timeline --days 7 --json --agent claude
cass timeline --since 7d --json

# Find related sessions for a file
cass context /path/to/source.ts --json

Status & Diagnostics

# Quick health (<50ms)
cass health
cass health --json

# Full status snapshot
cass status --json
cass state --json  # alias

# Statistics
cass stats --json
cass stats --by-source  # for multi-machine

# Full diagnostics
cass diag --verbose

---

Aggregation & Analytics

Aggregate search results server-side to get counts and distributions without transferring full result data:

# Count results by agent
cass search "error" --robot --aggregate agent
# → { "aggregations": { "agent": { "buckets": [{"key": "claude_code", "count": 45}, ...] } } }

# Multi-field aggregation
cass search "bug" --robot --aggregate agent,workspace,date

# Combine with filters
cass search "TODO" --agent claude --robot --aggregate workspace

Aggregation Field	Description
agent	Group by agent type (claude_code, codex, cursor, etc.)
workspace	Group by workspace/project path
date	Group by date (YYYY-MM-DD)
match_type	Group by match quality (exact, prefix, fuzzy)

Top 10 buckets returned per field, with other_count for remaining items.

---

Remote Sources (Multi-Machine Search)

Search across sessions from multiple machines via SSH/rsync.

Setup Wizard (Recommended)

cass sources setup

The wizard:

1. Discovers SSH hosts from ~/.ssh/config
2. Probes each for agent data and cass installation
3. Optionally installs cass on remotes
4. Indexes sessions on remotes
5. Configures sources.toml
6. Syncs data locally

cass sources setup --hosts css,csd,yto  # Specific hosts only
cass sources setup --dry-run             # Preview without changes
cass sources setup --resume              # Resume interrupted setup

Manual Setup

# Add a remote machine
cass sources add user@laptop.local --preset macos-defaults
cass sources add dev@workstation --path ~/.claude/projects --path ~/.codex/sessions

# List sources
cass sources list --json

# Sync sessions
cass sources sync
cass sources sync --source laptop --verbose

# Check connectivity
cass sources doctor
cass sources doctor --source laptop --json

# Path mappings (rewrite remote paths to local)
cass sources mappings list laptop
cass sources mappings add laptop --from /home/user/projects --to /Users/me/projects
cass sources mappings test laptop /home/user/projects/myapp/src/main.rs

# Remove source
cass sources remove laptop --purge -y

Configuration stored in ~/.config/cass/sources.toml (Linux) or ~/Library/Application Support/cass/sources.toml (macOS).

---

Robot Mode Deep Dive

Self-Documenting API

CASS teaches agents how to use itself:

# Quick capability check
cass capabilities --json
# Returns: features, connectors, limits

# Full API schema
cass introspect --json
# Returns: all commands, arguments, response shapes

# Topic-based docs (LLM-optimized)
cass robot-docs commands   # all commands and flags
cass robot-docs schemas    # response JSON schemas
cass robot-docs examples   # copy-paste invocations
cass robot-docs exit-codes # error handling
cass robot-docs guide      # quick-start walkthrough
cass robot-docs contracts  # API versioning
cass robot-docs sources    # remote sources guide

Forgiving Syntax (Agent-Friendly)

CASS auto-corrects common mistakes:

What you type	What CASS understands
cass serach "error"	cass search "error" (typo corrected)
cass -robot -limit=5	cass --robot --limit=5 (single-dash fixed)
cass --Robot --LIMIT 5	cass --robot --limit 5 (case normalized)
cass find "auth"	cass search "auth" (alias resolved)
cass --limt 5	cass --limit 5 (Levenshtein <=2)

Command Aliases:

• find, query, q, lookup, grep → search
• ls, list, info, summary → stats
• st, state → status
• reindex, idx, rebuild → index
• show, get, read → view
• docs, help-robot, robotdocs → robot-docs

Output Formats

# Pretty-printed JSON (default)
cass search "error" --robot

# Streaming JSONL (header + one hit per line)
cass search "error" --robot-format jsonl

# Compact single-line JSON
cass search "error" --robot-format compact

# With performance metadata
cass search "error" --robot --robot-meta

Design principle: stdout = JSON only; diagnostics go to stderr.

Token Budget Management

LLMs have context limits. Control output size:

Flag	Effect
--fields minimal	Only source_path, line_number, agent
--fields summary	Adds title, score
--fields score,title,snippet	Custom field selection
--max-content-length 500	Truncate long fields (UTF-8 safe)
--max-tokens 2000	Soft budget (~4 chars/token)
--limit 5	Cap number of results

Truncated fields include *_truncated: true indicator.

---

Structured Error Handling

Errors are JSON with actionable hints:

{
  "error": {
    "code": 3,
    "kind": "index_missing",
    "message": "Search index not found",
    "hint": "Run 'cass index --full' to build the index",
    "retryable": false
  }
}

Exit Codes

Code	Meaning	Action
0	Success	Parse stdout
1	Health check failed	Run cass index --full
2	Usage error	Fix syntax (hint provided)
3	Index/DB missing	Run cass index --full
4	Network error	Check connectivity
5	Data corruption	Run cass index --full --force-rebuild
6	Incompatible version	Update cass
7	Lock/busy	Retry later
8	Partial result	Increase --timeout
9	Unknown error	Check retryable flag

---

Search Modes

Three search modes, selectable with --mode flag:

Mode	Algorithm	Best For
lexical (default)	BM25 full-text	Exact term matching, code searches
semantic	Vector similarity	Conceptual queries, "find similar"
hybrid	Reciprocal Rank Fusion	Balanced precision and recall

cass search "authentication" --mode lexical --robot
cass search "how to handle user login" --mode semantic --robot
cass search "auth error handling" --mode hybrid --robot

Hybrid combines lexical and semantic using RRF:

RRF_score = Σ 1 / (60 + rank_i)

---

Pipeline Mode (Chained Search)

Chain searches by piping session paths:

# Find sessions mentioning "auth", then search within those for "token"
cass search "authentication" --robot-format sessions | \
  cass search "refresh token" --sessions-from - --robot

# Build a filtered corpus from today's work
cass search --today --robot-format sessions > today_sessions.txt
cass search "bug fix" --sessions-from today_sessions.txt --robot

Use cases:

• Drill-down: Broad search → narrow within results
• Cross-reference: Find sessions with term A, then find term B within them
• Corpus building: Save session lists for repeated searches

---

Query Language

Basic Queries

Query	Matches
error	Messages containing "error" (case-insensitive)
python error	Both "python" AND "error"
"authentication failed"	Exact phrase

Boolean Operators

Operator	Example	Meaning
AND	python AND error	Both terms required (default)
OR	error OR warning	Either term matches
NOT	error NOT test	First term, excluding second
-	error -test	Shorthand for NOT

# Complex boolean query
cass search "authentication AND (error OR failure) NOT test" --robot

# Exclude test files
cass search "bug fix -test -spec" --robot

# Either error type
cass search "TypeError OR ValueError" --robot

Wildcard Patterns

Pattern	Type	Performance
auth*	Prefix	Fast (edge n-grams)
*tion	Suffix	Slower (regex)
*config*	Substring	Slowest (regex)

Match Types

Results include match_type:

Type	Meaning	Score Boost
exact	Verbatim match	Highest
prefix	Via prefix expansion	High
suffix	Via suffix pattern	Medium
substring	Via substring pattern	Lower
fuzzy	Auto-fallback (sparse results)	Lowest

Auto-Fuzzy Fallback

When exact query returns <3 results, CASS automatically retries with wildcards:

• auth → *auth*
• Results flagged with wildcard_fallback: true

Flexible Time Input

CASS accepts a wide variety of time/date formats:

Format	Examples
Relative	-7d, -24h, -30m, -1w
Keywords	now, today, yesterday
ISO 8601	2024-11-25, 2024-11-25T14:30:00Z
US Dates	11/25/2024, 11-25-2024
Unix Timestamp	1732579200 (seconds or milliseconds)

---

Ranking Modes

Cycle with F12 in TUI or use --ranking flag:

Mode	Formula	Best For
Recent Heavy	relevance*0.3 + recency*0.7	"What was I working on?"
Balanced	relevance*0.5 + recency*0.5	General search
Relevance	relevance*0.8 + recency*0.2	"Best explanation of X"
Match Quality	Penalizes fuzzy matches	Precise technical searches
Date Newest	Pure chronological	Recent activity
Date Oldest	Reverse chronological	"When did I first..."

Score Components

• Text Relevance (BM25): Term frequency, inverse document frequency, length normalization
• Recency: Exponential decay (today ~1.0, last week ~0.7, last month ~0.3)
• Match Exactness: Exact phrase=1.0, Prefix=0.9, Suffix=0.8, Substring=0.6, Fuzzy=0.4

Blended Scoring Formula

Final_Score = BM25_Score × Match_Quality + α × Recency_Factor

Mode	α Value	Effect
Recent Heavy	1.0	Recency dominates
Balanced	0.4	Moderate recency boost
Relevance Heavy	0.1	BM25 dominates
Match Quality	0.0	Pure text matching

---

Supported Agents (11 Connectors)

Agent	Location	Format
Claude Code	~/.claude/projects	JSONL
Codex	~/.codex/sessions	JSONL (Rollout)
Gemini CLI	~/.gemini/tmp	JSON
Cline	VS Code global storage	Task directories
OpenCode	.opencode directories	SQLite
Amp	~/.local/share/amp + VS Code	Mixed
Cursor	~/Library/Application Support/Cursor	SQLite (state.vscdb)
ChatGPT	~/Library/Application Support/com.openai.chat	JSON (v1 unencrypted)
Aider	~/.aider.chat.history.md + per-project	Markdown
Pi-Agent	~/.pi/agent/sessions	JSONL with thinking
Factory (Droid)	~/.factory/sessions	JSONL by workspace

Note: ChatGPT v2/v3 are AES-256-GCM encrypted (keychain access required). Legacy v1 unencrypted conversations are indexed automatically.

---

TUI Features (for Humans)

Launch with cass (no flags):

Keyboard Shortcuts

Navigation:

• Up/Down: Move selection
• Left/Right: Switch panes
• Tab/Shift+Tab: Cycle focus
• Enter: Open in $EDITOR
• Space: Full-screen detail view
• Home/End: Jump to first/last result
• PageUp/PageDown: Scroll by page

Filtering:

• F3: Agent filter
• F4: Workspace filter
• F5/F6: Time filters (from/to)
• Shift+F3: Scope to current result's agent
• Shift+F4: Clear workspace filter
• Shift+F5: Cycle presets (24h/7d/30d/all)
• Ctrl+Del: Clear all filters

Modes:

• F2: Toggle theme (6 presets)
• F7: Context window size (S/M/L/XL)
• F9: Match mode (prefix/standard)
• F12: Ranking mode
• Ctrl+B: Toggle border style

Selection & Actions:

• m: Toggle selection
• Ctrl+A: Select all
• A: Bulk actions menu
• Ctrl+Enter: Add to queue
• Ctrl+O: Open all queued
• y: Copy path/content
• Ctrl+Y: Copy all selected
• /: Find in detail pane
• n/N: Next/prev match

Views & Palette:

• Ctrl+P: Command palette
• 1-9: Load saved view
• Shift+1-9: Save view to slot

Source Filtering (multi-machine):

• F11: Cycle source filter (all/local/remote)
• Shift+F11: Source selection menu

Global:

• Ctrl+C: Quit
• F1 or ?: Toggle help
• Ctrl+Shift+R: Force re-index
• Ctrl+Shift+Del: Reset all TUI state

Detail Pane Tabs

Tab	Content	Switch With
Messages	Full conversation with markdown	[ / ]
Snippets	Keyword-extracted summaries	[ / ]
Raw	Unformatted JSON/text	[ / ]

Context Window Sizing

Size	Characters	Use Case
Small	~200	Quick scanning
Medium	~400	Default balanced view
Large	~800	Longer passages
XLarge	~1600	Full context, code review

Peek Mode (Ctrl+Space): Temporarily expand to XL without changing default.

---

Theme Presets

Cycle through 6 built-in themes with F2:

Theme	Description	Best For
Dark	Tokyo Night-inspired deep blues	Low-light environments
Light	High-contrast light background	Bright environments
Catppuccin	Warm pastels, reduced eye strain	All-day coding
Dracula	Purple-accented dark theme	Popular developer theme
Nord	Arctic-inspired cool tones	Calm, focused work
High Contrast	Maximum readability	Accessibility needs

All themes validated against WCAG contrast requirements (4.5:1 minimum for text).

Role-Aware Message Styling

Role	Visual Treatment
User	Blue-tinted background, bold
Assistant	Green-tinted background
System	Gray/muted background
Tool	Orange-tinted background

---

Saved Views

Save filter configurations to 9 slots for instant recall.

What Gets Saved:

• Active filters (agent, workspace, time range)
• Current ranking mode
• The search query

Keyboard:

• Shift+1 through Shift+9: Save current view
• 1 through 9: Load view from slot

Via Command Palette: Ctrl+P → "Save/Load view"

Views persist in tui_state.json across sessions.

---

Density Modes

Control lines per search result. Cycle with Shift+D:

Mode	Lines	Best For
Compact	3	Maximum results visible
Cozy	5	Balanced view (default)
Spacious	8	Detailed preview

---

Bookmark System

Save important results with notes and tags:

In TUI: Press b to bookmark, add notes and tags.

Bookmark Structure:

• title: Short description
• source_path, line_number, agent, workspace
• note: Your annotations
• tags: Comma-separated labels
• snippet: Extracted content

Storage: ~/.local/share/coding-agent-search/bookmarks.db (SQLite)

---

Optional Semantic Search

Local-only semantic search using MiniLM (no cloud):

Required files (place in data directory):

• model.onnx
• tokenizer.json
• config.json
• special_tokens_map.json
• tokenizer_config.json

Vector index stored as vector_index/index-minilm-384.cvvi.

CASS does NOT auto-download models; you must manually install them.

Hash Embedder Fallback: When MiniLM not installed, CASS uses a hash-based embedder for approximate semantic similarity.

---

Watch Mode

Real-time index updates:

cass index --watch

• Debounce: 2 seconds (wait for burst to settle)
• Max wait: 5 seconds (force flush during continuous activity)
• Incremental: Only re-scans modified files

TUI automatically starts watch mode in background.

---

Deduplication Strategy

CASS uses multi-layer deduplication:

1. Message Hash: SHA-256 of (role + content + timestamp) - identical messages stored once
2. Conversation Fingerprint: Hash of first N message hashes - detects duplicate files
3. Search-Time Dedup: Results deduplicated by content similarity

Noise Filtering:

• Empty messages and pure whitespace
• System prompts (unless searching for them)
• Repeated tool acknowledgments

---

Performance Characteristics

Operation	Latency
Prefix search (cached)	2-8ms
Prefix search (cold)	40-60ms
Substring search	80-200ms
Full reindex	5-30s
Incremental reindex	50-500ms
Health check	<50ms

Memory: 70-140MB typical (50K messages) Disk: ~600 bytes/message (including n-gram overhead)

---

Response Shapes

Search Response:

{
  "query": "error",
  "limit": 10,
  "count": 5,
  "total_matches": 42,
  "hits": [
    {
      "source_path": "/path/to/session.jsonl",
      "line_number": 123,
      "agent": "claude_code",
      "workspace": "/projects/myapp",
      "title": "Authentication debugging",
      "snippet": "The error occurs when...",
      "score": 0.85,
      "match_type": "exact",
      "created_at": "2024-01-15T10:30:00Z"
    }
  ],
  "_meta": {
    "elapsed_ms": 12,
    "cache_hit": true,
    "wildcard_fallback": false,
    "next_cursor": "eyJ...",
    "index_freshness": { "stale": false, "age_seconds": 120 }
  }
}

Aggregation Response:

{
  "aggregations": {
    "agent": {
      "buckets": [
        {"key": "claude_code", "count": 120},
        {"key": "codex", "count": 85}
      ],
      "other_count": 15
    }
  }
}

---

Environment Variables

Variable	Purpose
CASS_DATA_DIR	Override data directory
CHATGPT_ENCRYPTION_KEY	Base64 key for encrypted ChatGPT
PI_CODING_AGENT_DIR	Override Pi-Agent sessions path
CASS_CACHE_SHARD_CAP	Per-shard cache entries (default 256)
CASS_CACHE_TOTAL_CAP	Total cached hits (default 2048)
CASS_DEBUG_CACHE_METRICS	Enable cache debug logging
CODING_AGENT_SEARCH_NO_UPDATE_PROMPT	Skip update checks

---

Shell Completions

cass completions bash > ~/.local/share/bash-completion/completions/cass
cass completions zsh > "${fpath[1]}/_cass"
cass completions fish > ~/.config/fish/completions/cass.fish
cass completions powershell >> $PROFILE

---

API Contract & Versioning

cass api-version --json
# → { "version": "0.4.0", "contract_version": "1", "breaking_changes": [] }

cass introspect --json
# → Full schema: all commands, arguments, response types

Guaranteed Stable:

• Exit codes and their meanings
• JSON response structure for --robot output
• Flag names and behaviors
• _meta block format

---

Integration with CASS Memory (cm)

CASS provides episodic memory (raw sessions). CM extracts procedural memory (rules and playbooks):

# 1. CASS indexes raw sessions
cass index --full

# 2. Search for relevant past experience
cass search "authentication timeout" --robot --limit 10

# 3. CM reflects on sessions to extract rules
cm reflect

---

Troubleshooting

Issue	Solution
"missing index"	cass index --full
Stale warning	Rerun index or enable watch
Empty results	Check cass stats --json, verify connectors detected
JSON parsing errors	Use --robot-format compact
Watch not triggering	Check watch_state.json, verify file event support
Reset TUI state	cass tui --reset-state or Ctrl+Shift+Del

---

Installation

# One-liner install
curl -fsSL https://raw.githubusercontent.com/Dicklesworthstone/coding_agent_session_search/main/install.sh \
  | bash -s -- --easy-mode --verify

# Windows
irm https://raw.githubusercontent.com/Dicklesworthstone/coding_agent_session_search/main/install.ps1 | iex

---

Integration with Flywheel

Tool	Integration
CM	CASS provides episodic memory, CM extracts procedural memory
NTM	Robot mode flags for searching past sessions
Agent Mail	Search threads across agent history
BV	Cross-reference beads with past solutions