AICTX
Quickstart

Try AICTX in 60 seconds.

Install once, initialize the repo, then let supported agents resume before work and finalize evidence after work.

The whole product is built around one loop:

resume before work -> work normally -> finalize evidence -> next session continues
install → init → use your agent 
pip install aictx
aictx install
aictx init

# Then keep using your coding agent.
# Supported agents are instructed to:
# resume context → work → finalize continuity

You install and initialize AICTX. Supported agents are instructed to resume before work and finalize after work. Everything is stored locally under .aictx/.

Optional check:

aictx --version
aictx doctor --repo . --json

Fresh repos start with little continuity

A new repo may not have much to resume yet. That is expected: AICTX becomes useful as agents finalize work and leave Work State, failures, decisions, or handoffs behind.

MCP support

By default, aictx install prepares AICTX MCP runtime metadata and aictx init writes repo-local MCP config for compatible clients.

Launch local stdio server

aictx mcp-server --repo . --profile full

Inspect MCP setup

aictx mcp status --repo .

Opt out

aictx install --no-mcp
aictx init --no-mcp

Agents should prefer MCP tools such as aictx_resume, aictx_finalize, and aictx_view when available, and fall back to CLI commands otherwise.

Interactive install

Installation prepares AICTX MCP support by default and will only ask you about enabling recommended RepoMap support using Tree-sitter.

RepoMap uses Tree-sitter to build a compact structural map of files and symbols. It helps agents choose better starting points without reading the whole repo.

2. Normal flow example

  • Session starts
  • The user asks for a task
  • AICTX provides the agent with a continuity resume capsule
codex@my-repo · session #12 · awake

Resuming: parser refactor was paused after updating token tests.
Last progress: `tests/test_parser.py` passes; next step is to update error recovery cases.

────────────────────────────────
  • Agent performs the task
  • AICTX updates continuity artifacts
  • Agent provides the user with a continuity summary
────────────────────────────────
AICTX summary

Context: resumed parser refactor from previous session state.
Map: RepoMap quick ok.
Saved: updated handoff and continuity state.
Validation: `pytest -q tests/test_parser.py` passed.
Next: update parser error recovery cases.
Details: last_execution_summary.md
Continuity view file: continuity-map.mmd
View continuity online: mermaid.live view

3. Inspect Continuity View Manually

aictx view --repo .

Default output:

.aictx/reports/continuity-view.md
.aictx/reports/continuity-map.mmd

Not hidden memory. Reviewable operational continuity.

Agent lifecycle in one minute

The lifecycle is intentionally small: resume a useful brief, work normally, then finalize factual evidence for the next session.

  • resume --brief can return a smaller startup payload.
  • Supported integrations get compact runner contracts and guard triggers.
  • Validation expectations are task-aware, so documentation and analysis tasks stay advisory.
  • finalize can surface dirty edited files without staging or committing them.
  • Work State can preserve compact discarded hypotheses when an agent explicitly records an abandoned approach.
  • Prior successful strategies can be used internally as bounded hints when they are relevant.

Next steps

User Manual

Copy-paste commands for setup, inspection, MCP, portability, reports and cleanup.

Open User Manual →

Technical Documentation

Architecture and reference docs for Work State, contracts, RepoMap, MCP and continuity artifacts.

Open Technical Documentation →