jsat 0.1.7

Codebase intelligence shell and SDK — index once, query forever, works with Claude, GPT, Ollama, and any AI

JSAT — JaySoft AI Tools

Codebase intelligence for AI sessions — index once, query forever, works with any AI.

---

What is JSAT?

Every AI session starts with the same problem: you spend the first ten minutes re-explaining your architecture, re-pasting function signatures, and re-describing how services talk to each other. JSAT solves this by building a persistent graph of your codebase once — functions, classes, files, services, API endpoints, database tables, Kafka topics, and every relationship between them — and making that context instantly available to any AI you use.

JSAT works as a CLI, a Python SDK, and an MCP server that plugs directly into Claude Code. If Claude Code CLI is installed, JSAT uses it automatically with no API key required. For everything else — Anthropic API, OpenAI, Gemini, Ollama, LM Studio — one command switches the provider.

---

Quick Start

pip install jsat

# Wire JSAT into Claude Code (if installed — no API key needed)
jsat connect claude

# Index your project
cd your-project/
jsat index .

# Open Claude with full codebase context and JSAT tools available
jsat claude

Inside Claude Code you can then use slash commands:

/jsat-query what does the payment service do?
/jsat-blast-radius src/payment/refund.py
/jsat-security
/jsat-incident "500 errors spiking on checkout"

---

Installation

JSAT ships as a minimal core with optional extras. Install only what you need.

Extra	What's added	Approx. size	When to use
(none) / core	tree-sitter parsers, SQLite graph, CLI	~80 MB	Starting point for any setup
local	Ollama client	+small	Local models via Ollama
standard	Semgrep, OpenAPI/AsyncAPI validator, more language parsers (Java, Ruby, Rust)	+medium	Security reviews, API contract checks
team	Neo4j, Qdrant, Redis, Graphiti (includes standard)	+large	Shared graph across a team
anthropic	Anthropic Python SDK	+small	Claude API (key required)
openai	OpenAI Python SDK	+small	GPT-4o, GPT-4o-mini (key required)
ci	PyGitHub, SARIF tools (includes standard)	+small	CI/CD pipelines, GitHub Actions
all	Everything above	+large	Full feature set

pip install jsat                   # core only
pip install 'jsat[local]'          # + Ollama
pip install 'jsat[standard]'       # + security analysis, OpenAPI validation
pip install 'jsat[team]'           # + Neo4j, Qdrant, Redis
pip install 'jsat[anthropic]'      # + Claude API SDK
pip install 'jsat[openai]'         # + OpenAI SDK
pip install 'jsat[all]'            # everything

---

AI Providers

JSAT auto-detects available providers at startup and picks the best one in priority order:

1. Claude Code CLI — detected via which claude; no API key, no extra SDK, full tool calling
2. Anthropic API — if ANTHROPIC_API_KEY is set and jsat[anthropic] is installed
3. OpenAI — if OPENAI_API_KEY is set and jsat[openai] is installed
4. Ollama — if ollama serve is running at localhost:11434
5. LM Studio — if an OpenAI-compatible server is running at localhost:1234
6. No AI — tools that don't need AI (indexing, blast radius, export) still work

Check what's available

jsat ai status        # shows all providers, which is active, and switch commands

Switch providers

jsat ai use ollama                        # free, local, no key
jsat ai use ollama --model qwen2.5-coder:7b
jsat ai use anthropic                     # needs ANTHROPIC_API_KEY
jsat ai use openai --model gpt-4o-mini    # needs OPENAI_API_KEY
jsat ai use lmstudio                      # any OpenAI-compat server at localhost:1234
jsat ai test                              # verify the configured provider works

Switch inside the JSAT shell

switch claude    → Claude Code CLI (no key) or Claude API
switch gpt       → GPT-4o
switch ollama    → local Ollama
switch haiku     → Claude Haiku
switch phi       → phi3:mini (fast, low RAM)
switch lmstudio  → LM Studio

---

Claude Code Integration

JSAT's tightest integration is with Claude Code. One command registers JSAT as an MCP server and installs /jsat-* slash commands. After that, Claude can call JSAT tools automatically — or you can invoke them explicitly.

Setup

jsat connect claude                       # project-level (this repo only)
jsat connect claude --scope global        # global (all Claude Code sessions)

Restart Claude Code. JSAT tools are now available.

Open Claude with JSAT context pre-loaded

jsat claude

Slash commands installed into Claude Code

Command	What it does
/jsat-query <question>	Natural language query over the indexed graph
/jsat-blast-radius <file or symbol>	Trace downstream impact grouped by severity
/jsat-security [path]	Security scan — Critical and High issues first
/jsat-incident <description>	Root-cause hypotheses ranked by confidence
/jsat-index [path]	Rebuild the codebase graph
/jsat-status	Node and edge counts
/jsat-doctor	Full health check
/jsat-ithinking <task>	Run the IThinking structured planning framework for a task
/jsat-think <task>	Alias for /jsat-ithinking — shorthand for quick invocation
/jsat-prompt-diff <query>	Show raw input vs optimized prompt side by side

MCP tools Claude can call automatically

JSAT defines 47 MCP tool slots across 12 categories; 38 are currently wired and active in the MCP server. Highlights:

• jsat__query — answer any codebase question
• jsat__blast_radius_file, jsat__blast_radius_diff, jsat__blast_radius_symbol, jsat__blast_radius_topic
• jsat__security_scan_file, jsat__get_auth_coverage, jsat__list_secrets, jsat__get_dependency_cves
• jsat__investigate_incident, jsat__generate_runbook
• jsat__check_breaking_changes, jsat__get_compat_score
• jsat__validate_migration, jsat__suggest_zero_downtime
• jsat__submit_for_review, jsat__get_high_confidence_bugs
• jsat__ithinking_plan, jsat__ithinking_execute
• jsat__prompt_optimize, jsat__prompt_diff

Connect to Cursor

jsat connect cursor        # writes to ~/.cursor/mcp.json

---

Prompt Optimizer

JSAT automatically optimizes every query through a 7-stage pipeline before sending it to the AI:

# Print the optimized prompt (inspect mode)
jsat prompt "improve the retry logic"

# Optimize + send to AI
jsat prompt --send "improve the retry logic"

# See what you typed vs what the AI actually received
jsat prompt --diff "improve the retry logic"

# Format, CoT, specific AI
jsat prompt --send --format code --ai claude "write a test for refund()"

Stages:

1. Task classification (code_gen / refactor / review / debug / question / plan / test / security)
2. Context injection from the JSAT graph (BFS, 70/30 recency split)
3. Constraint injection from the knowledge base (ADRs, coding standards)
4. Few-shot example selection (kNN over prompt history)
5. Output format specification (code only / JSON findings / prose / numbered steps)
6. Model-specific formatting (XML for Claude, Markdown for GPT, plain for Ollama)
7. Token compression (example shortening → block removal → docstring stripping)

In the shell — every message is auto-optimized:

jsat [Claude Code (CLI)]> improve the retry logic

✦ Optimized refactor | 6→847 tokens (35% saved) | 3 ctx nodes | opt show to see diff

Claude: Here's the improved retry using tenacity...

Shell commands:

opt on        # enable auto-optimization (default)
opt off       # disable for the current session
opt show      # show raw input vs full optimized prompt side by side
opt history   # browse past optimization diffs

See before/after:

jsat> opt show

Shows both panels: raw input vs full optimized prompt with injected context, constraints, and model formatting.

Disconnect or remove

jsat disconnect claude                        # project-level
jsat disconnect claude --scope global         # global
jsat disconnect claude --scope all            # everywhere
jsat remove                                   # remove all JSAT artifacts from this repo

---

CLI Reference

Core commands

Command	Description
jsat index [path]	Build or update the codebase graph
jsat index . --force	Full re-index (skip incremental check)
jsat index . --languages python,go	Index specific languages only
jsat shell	Start the interactive JSAT REPL
jsat claude	Open Claude Code with JSAT MCP tools loaded
jsat gpt	Open a GPT session with JSAT tools
jsat ollama [--model llama3.2]	Open a local Ollama session
jsat prompt <query>	Print the optimized prompt (inspect without sending)
jsat prompt --send <query>	Optimize prompt and send to the configured AI
jsat prompt --diff <query>	Show raw input vs optimized prompt side by side
jsat prompt --format code|plan|json|prose	Override output format for this prompt
jsat prompt --ai claude|gpt|ollama	Override AI provider for this prompt
jsat doctor	System health check (graph, AI, services)
jsat doctor --json	Health check as raw JSON
jsat version	Print JSAT version

Configuration

Command	Description
jsat init	Generate .jsat/config.yaml (default: solo profile)
jsat init --profile team	Team profile (Neo4j, Qdrant, Redis, Claude API)
jsat init --profile ci	CI profile (SQLite, no AI, JSON logs)
jsat init --profile raspberry-pi	Low-RAM profile (SQLite, phi3:mini, batch size 8)

AI provider management

Command	Description
jsat ai status	Show all providers: available, active, free/paid
jsat ai use <provider>	Configure a provider and write to config
jsat ai use ollama --model phi3:mini	Use a specific Ollama model
jsat ai test	Send a test prompt and verify the provider works
jsat ai models	List models available from the configured provider

Claude Code integration

Command	Description
jsat connect claude	Wire JSAT into Claude Code (project scope)
jsat connect claude --scope global	Wire JSAT globally (all sessions)
jsat connect claude --no-skills	MCP only — skip installing slash commands
jsat connect cursor	Wire JSAT into Cursor
jsat connect list	Show all active JSAT MCP configs
jsat disconnect claude	Remove from project Claude Code config
jsat disconnect claude --scope all	Remove from all scopes

Export and import

Command	Description
jsat export backup.jsat.zip	Export the current index as a portable zip
jsat export backup.jsat.zip -z 9	Export with maximum compression
jsat import backup.jsat.zip	Restore an index from an exported archive

Skills

Command	Description
jsat skills list	List installed JSAT skill manifests
jsat skills run <name>	Run a named skill with optional key=val args
jsat ci-setup	Write a GitHub Actions workflow for JSAT
jsat ci-setup --provider gitlab	Write a GitLab CI pipeline for JSAT

---

Python SDK

from jsat import JSAT

# Instantiate — auto-detects AI provider, loads config
js = JSAT(repo=".")

# Build the graph (incremental by default)
result = js.index()
print(f"Indexed {result.nodes_indexed} nodes, {result.edges_indexed} edges")

# Natural language query over the graph
result = js.query("what calls the refund endpoint?")
print(result.answer)

# Trace blast radius of a change
report = js.blast_radius("src/payment/refund.py")
for item in report.impacts:
    print(f"{item.severity:10s}  {item.node_id}")

# Security analysis (requires jsat[standard])
sec = js.security_review(path="src/")
for finding in sec.findings:
    print(f"{finding.severity}: {finding.title} — {finding.file}:{finding.line}")

# Incident investigation
incident = js.investigate_incident("500 errors on checkout", since="24h")
for h in incident.hypotheses:
    print(f"[{h.score:.0%}] {h.title}")

# Export the index for sharing or CI caching
manifest = js.export("snapshot.jsat.zip")
print(f"Exported {manifest.size_mb:.1f} MB")

# Restore from an export
js2 = JSAT.from_import("snapshot.jsat.zip")

# Switch AI provider mid-session
js.switch_ai("ollama", model="qwen2.5-coder:7b")
js.switch_ai("anthropic")
js.switch_ai("gpt", model="gpt-4o-mini")

# Health check
health = js.doctor()
print(health["profile"], health["graph"]["backend"])

---

Tools Overview

#	Tool	Description
0	JSAT Shell	Interactive REPL — run any tool directly, switch AI mid-session, no AI required
1	Directory Indexer	Walks a repo, parses source with tree-sitter, writes nodes and edges to the graph
2	Test Intelligence Helper	Finds test gaps, maps behaviors to coverage, generates unit/integration/contract tests
3	Feature Helper	Answers "how do I add X?" using graph context — finds relevant files and patterns
4	Blast Radius Analyzer	BFS over the graph to trace downstream impact; classifies edges as breaking/degraded/warning/safe
5	API Contract Validator	Diffs OpenAPI/AsyncAPI specs, classifies breaking changes, scores backward compatibility (0–100)
6	Security Review Agent	OWASP pattern scan, auth coverage gaps, hardcoded secret detection, dependency CVE lookup
7	Incident Investigation Helper	Correlates an incident description against recent commits and graph topology; ranks root-cause hypotheses
8	Migration Safety Validator	Validates migration files, estimates lock duration, generates zero-downtime migration plans
9	Multi-Model Code Review (true parallel dispatch)	Dispatches a diff to multiple AI models simultaneously via ThreadPoolExecutor; surfaces only bugs confirmed by two or more models
10	Knowledge Base Builder	Persistent searchable store of architectural decisions, runbooks, and tribal knowledge
11	Multi-Agent Orchestrator	Decomposes a task and runs specialized sub-agents (understanding, generation, review, test, security, docs)
12	Export / Import System	Portable zip snapshots of the full graph — share between machines, cache in CI, restore in seconds
13	Python SDK	Programmatic access to every tool via from jsat import JSAT
14	IThinking Meta-Cognitive Layer	Structured seven-phase reasoning: clarify, plan, context, assumptions, execute, reflect — with human approval gates

Multi-Model Review

Tool 9 dispatches the diff to all configured models in parallel, collects findings, and surfaces only those confirmed by two or more models. Configure the model list and timeout in .jsat/config.yaml:

review:
  models:
    - {provider: claude_cli, model: claude-sonnet-4-6}
    - {provider: ollama, model: qwen2.5-coder:7b}
  parallel_timeout_seconds: 90
  min_confidence: medium

• parallel_timeout_seconds — per-review wall-clock deadline; any model that exceeds this is skipped and its absence is logged.
• min_confidence — minimum agreement level to surface a finding: low (any model), medium (2+ models), high (all models).

---

Graph Schema

JSAT indexes these node types and relationship edges:

Nodes: function, class, file, service, endpoint, table, kafka_topic

Edges:

Edge	Meaning
CALLS	Function A calls function B
IMPORTS	File A imports module B
READS_FROM	Code reads from a table or topic
WRITES_TO	Code writes to a table or topic
PRODUCES	Service produces a Kafka message
CONSUMES	Service consumes a Kafka topic
IMPLEMENTS	Class implements an interface
INHERITS	Class inherits from another
DEPENDS_ON	Service depends on another service

---

Configuration

JSAT stores all state under .jsat/ in your repo root. The config file is .jsat/config.yaml.

jsat init                        # write starter config (solo profile)
jsat init --profile team         # team profile
jsat init --profile ci           # CI/CD profile

Key settings in .jsat/config.yaml:

graph:
  backend: sqlite          # sqlite (default) | neo4j (team profile)
  path: .jsat/graph/graph.db

embeddings:
  provider: local          # local | openai | none
  model: nomic-embed-code

ai:
  provider: ollama         # ollama | anthropic | openai | openai_compat | claude_cli | none
  model: llama3.2
  base_url: null           # for openai_compat (LM Studio, Gemini, etc.)

cache:
  backend: memory          # memory | disk | redis

indexer:
  languages: [python, javascript, go, java, ruby, rust]
  exclude_patterns: ["**/node_modules/**", "**/.git/**", "**/dist/**"]
  max_file_size_kb: 500

ithinking:
  enabled: true
  mode: interactive        # interactive | silent
  gate_level: medium       # low | medium | high

Profiles at a glance

Profile	Graph	AI	Cache	Use case
solo	SQLite	Ollama / llama3.2	Memory	Individual developer, no external services
team	Neo4j	Claude API	Redis	Shared graph, team-wide knowledge base
ci	SQLite	None	Memory	GitHub Actions, no API keys, JSON logs
raspberry-pi	SQLite	Ollama / phi3:mini	Disk	Low-RAM devices, batch size 8

Config search order

JSAT finds its config by checking these locations in order (first found wins):

1. Explicit path passed to JSAT(config=...) or --config flag
2. $JSAT_CONFIG environment variable
3. {repo}/.jsat/config.yaml (canonical)
4. {repo}/.jsat.yaml (legacy)
5. ./.jsat/config.yaml (CWD)
6. ~/.config/jsat/config.yaml
7. /etc/jsat/config.yaml

---

Supported Languages

Language	Parser	Notes
Python	tree-sitter-python	Core (always available)
JavaScript / TypeScript	tree-sitter-javascript	Core
Go	tree-sitter-go	Core
Java	tree-sitter-java	Requires jsat[standard]
Ruby	tree-sitter-ruby	Requires jsat[standard]
Rust	tree-sitter-rust	Requires jsat[standard]

---

Contributing

Contributions are welcome. Please open an issue or pull request on GitHub.

• Repository: github.com/iamjpsonkar/JaySoft-AI_Tools
• Bug reports: open an issue on GitHub
• Author: Jay Prakash Sonkar — [e-mail]
• License: MIT

---

License

MIT License. Copyright (c) Jay Prakash Sonkar.