About Bitloops

Bitloops is a memory and context layer for AI coding agents.

AI agents can generate code quickly, but the reasoning behind their changes is
often lost. Sessions restart from zero, context drifts across tools, and
reviewers are left with diffs that do not explain how a decision was made.

Bitloops captures and structures agent reasoning alongside your repository. It
links conversations, decisions, and generated code back to commits so teams can
understand, review, and improve AI-assisted development.

Instead of sending large amounts of repository context to the model, Bitloops
retrieves the most relevant code, architecture, and prior reasoning. This gives
agents higher-quality context, reduces token usage, and improves the likelihood
of correct results on the first attempt.

---

Why Bitloops

Bitloops introduces three core capabilities:

• Repository memory for AI agents
• Targeted context retrieval for codebases
• Traceable AI reasoning through Git commits

Together, these allow agents to work with better context while teams retain
visibility and governance over AI-generated changes.

Without Bitloops	With Bitloops
Each session starts from zero	Sessions build on shared repository memory
Agents send large amounts of repository context to the model	Only relevant code, architecture, and reasoning are retrieved
High token usage with noisy context	Fewer tokens with higher-quality context
Reviewers see diffs with no reasoning trail	Commits link to the developer–AI conversation
AI reasoning disappears between sessions	Agent reasoning remains searchable
Limited governance over AI-generated changes	Teams can review and audit AI reasoning through commits
Multiple iterations to reach a correct result	Higher chance of getting it right the first time

---

What Bitloops does

Bitloops adds structured memory, context retrieval, and reasoning traceability
to AI-assisted development.

• Shared memory for coding agents: Supported tools access the same repository-scoped knowledge layer, turning
  context into shared infrastructure instead of siloed sessions.

• Git-linked reasoning capture: Bitloops records the developer–agent workflow around each commit so teams can
  trace how a change was produced, not just what appears in the diff.

• Targeted context retrieval: Instead of forcing every session to search the codebase from scratch, Bitloops
  retrieves the most relevant structural and historical signals for the task.

• Local observability: A local dashboard lets teams inspect AI-assisted activity without sending code
  or commit history to a cloud service.

• External knowledge integration: Connect GitHub issues, pull requests, Jira tickets, and Confluence pages
  directly to your repository context.

Installation

Native Install (Recommended)

macOS, Linux, WSL:

curl -fsSL https://bitloops.com/install.sh | bash

Windows PowerShell:

irm https://bitloops.com/install.ps1 | iex

Windows CMD:

curl -fsSL https://bitloops.com/install.cmd -o install.cmd && install.cmd && del install.cmd

Homebrew (macOS/Linux)

brew install bitloops/tap/bitloops

Getting Started

1. Start the daemon and create the global daemon config:

   bitloops start --create-default-config
   

   On a fresh machine, interactive bitloops start can also prompt to create the default daemon config. During that first bootstrap, Bitloops asks for telemetry consent unless you pass --telemetry, --telemetry=false, or --no-telemetry.

2. From inside the git repository or subproject you want to capture, bootstrap Bitloops:

   bitloops init
   

3. Toggle capture later if needed:

   bitloops enable
   bitloops disable
   

4. Work as usual and commit normally. Bitloops will capture the relevant
   developer-agent context around those changes.

Dashboard

To view your checkpoints, run the following command again from within the root of your repo:

bitloops dashboard

To control the daemon directly:

bitloops start
bitloops daemon stop
bitloops status
bitloops checkpoints status

Uninstall

Remove Bitloops hooks from the current repository:

bitloops disable

Remove Bitloops-managed artefacts from your machine as well:

bitloops uninstall --full

Supported Agents

• Claude Code
• Codex (currently supports SessionStart and Stop hooks only; richer hook parity will follow as Codex expands hook coverage)
• Cursor
• Gemini
• Copilot
• OpenCode

What is DevQL?

DevQL is a typed GraphQL interface for querying artefacts, checkpoints, dependencies, and knowledge — available as a CLI DSL, raw GraphQL, or dashboard endpoint.

Read more here

External Knowledge

Bitloops can ingest repository-scoped external knowledge by URL:

bitloops devql knowledge add https://github.com/bitloops/bitloops/issues/42
bitloops devql knowledge add https://bitloops.atlassian.net/browse/CLI-1370 --commit <sha>

Supported sources:

• GitHub issues
• GitHub pull requests
• Jira issues
• Confluence pages

For this flow, SQLite stores repository-scoped identity and relations, DuckDB stores version metadata, and the full payload content is stored in the configured blob backend.

FAQs

Do you need access to my codebase?

No. None of your code is sent to our servers. By default Bitloops keeps its
config, cache, state, and local data in platform app directories on your
machine, with store backends controlled by the daemon config.

Is this totally free for real?

You bet!

What kind of databases do I need?

DevQL now uses a provider model:

• Relational backend: sqlite or postgres
• Events backend: duckdb or clickhouse

Current runtime adapters are sqlite/postgres for relational and
duckdb/clickhouse for events (default relational backend: sqlite,
default events backend: duckdb). Legacy postgres_dsn / clickhouse_* and
BITLOOPS_DEVQL_* settings remain supported for backward compatibility.

Why do you use telemetry and why should I opt-in?

Telemetry data help us understand which features users are using the most and
help us guide our development. The telemetry data are not connected to specific
users and are analysed and considered in aggregate. Bitloops asks for consent
when the default daemon config is first created, and later interactive init
or enable runs only ask again if consent becomes unresolved for an existing
daemon config.

Community & Support

Contributing

We welcome contributions from the community! Your input helps make Bitloops better
for everyone. See CONTRIBUTING.md to get started.

We welcome:

• bug reports
• documentation improvements
• performance improvements
• new integrations with AI coding tools
• architectural ideas and discussions

If you are unsure where to start, open an issue and we will help guide you.

Code of Conduct

We're committed to fostering an inclusive and respectful community. Read our
CODE_OF_CONDUCT.md for guidelines.

License

Apache 2.0.