Agent sessions stream in
Non-blocking hooks for Codex and Claude Code send prompts, tools, and outcomes into a private workspace with no change to how engineers work.
Workflow memory for AI engineering teams
Stop re-solving the same AI work
Tacit captures every Codex and Claude Code session, detects the patterns your team keeps rediscovering, and turns them into reusable loop specs anyone can run.
Install after sign-in
npx @alekskapera/tacitCodex & Claude Code hooks · Private raw traces · Shared loop specs
Built into the tools your team already runs
OpenAI CodexClaude CodeMarkdown loop specs
How it works
The work your team repeats becomes operational knowledge.
Repeated work surfaces
Tacit segments sessions into episodes, clusters similar intent and outcomes, and highlights the loops your team keeps rediscovering.
Loops become reusable
Turn private draft candidates into Markdown loop specs the whole team can copy, download, and run as prompt-ready playbooks.
What a loop looks like
A loop spec is a playbook your next session can run.
Every published loop exposes when to use it, the required inputs, agent instructions, success criteria, and the evidence behind it as plain Markdown, ready to paste into the next agent session.
loop-spec.md
# Loop: migrate-prisma-schemawhen_to_use: Adding or changing a Prisma modelseen: 7 sessions across 3 engineers## inputs- model name and fields- target migration env## agent_steps1. edit schema.prisma2. run `prisma migrate dev`3. update affected service typessuccess: tests pass, no drift
Private by default
Raw sessions stay private. Useful loops can become shared.
Your trace is yours
Raw prompts, outputs, imports, events, episodes, and private drafts stay attached to the originating user until they choose to publish.
The reusable part is shared
Published loops expose only the playbook: when to use it, required inputs, agent instructions, success criteria, evidence, and risks.
Pricing
Credit plans for captured AI work.
One credit roughly covers one message. Credits apply to accepted hook uploads and saved imports.
Free
Available$0
10,000 credits / month
For individuals testing capture, imports, and loop publishing on a small local workflow.
Start freePlus
Coming soon$19/month
100,000 credits / month
For active AI engineering teams that want more capture volume and faster review cycles.
Coming soonEnterprise
Coming soonCustom
Custom credits
For organizations that need custom credit volume, deployment support, and retention policies.
Coming soonQuestions
What teams ask before they install.
Will the hooks slow down my agent?
No. Capture runs through non-blocking hooks that hand sessions off asynchronously. Your Codex and Claude Code runs behave the way they do today.
Is my code and prompt data private?
Yes, by default. Raw prompts, outputs, imports, events, and private drafts stay attached to the user who created them.
Which tools does Tacit work with?
Tacit captures Codex and Claude Code sessions today via lightweight CLI hooks. You install once with a single command and keep working normally.
What exactly is a loop spec?
A loop spec is a reusable Markdown playbook Tacit generates from work your team repeats: when to use it, inputs, agent steps, success criteria, and risks.
Do I have to change how my team works?
No new workflow to adopt. Engineers keep using Codex and Claude Code the way they already do while Tacit watches for repeated patterns.
What does it cost to start?
Free includes 10,000 credits a month, enough to install the hooks, capture real sessions, and publish your first reusable loop.
Still deciding? The fastest answer is to capture a few of your own sessions.
Start freeStart with the loops your team is already creating.
Install the hooks, capture a few sessions, and watch your first reusable loops surface with Free credits included.
Start free