Intent, decisions, blockers — pulled from every Claude Code, Codex, and Gemini session

You think with Claude Codex Gemini .
Tacit keeps the log.

"What did I decide about auth three weeks ago?" Open Tacit — every session broken down, searchable by project.

BUY TACIT — $29 → TRY FREE →

My own usage: 88 projects · 644 sessions · 150 days in

100% local No cloud One-time purchase Claude Code Codex Gemini Mac Windows Linux
Claude Code · Codex · Gemini · runs locally
What your agent sessions look like — before vs after Tacit On the left, a terminal showing raw JSONL output from a 454-message Claude Code, Codex, or Gemini session, 2.3 megabytes of unreadable JSON. An orange arrow labeled "~2ms" points to the right, where Tacit has extracted the same session into a structured report titled "Refactor store pattern for archive module" with five sections: Intent, Decisions, Blockers, Outcomes, and a Phases timeline. What Your Agent Sessions Look Like Before vs After Tacit BEFORE session.jsonl {"type":"assistant","message":{"id":"msg_ 01XYz","type":"message","role":"assistant ","content":[{"type":"text","text":"Let m e check the existing implementation first "}],"model":"claude-opus-4-6","stop_reas on":"end_turn","usage":{"input_tokens":1 2847,"output_tokens":324}},"parentUuid": "abc-123","uuid":"def-456","timestamp":"2 026-04-08T10:15:23Z"} {"type":"human","message":{"id":"msg_02A Bc","type":"message","role":"user","conte nt":[{"type":"text","text":"No, use the e xisting store pattern instead. We decided against a new abstraction last week."}]}} {"type":"assistant","message":{"id":"msg_ 03DEf","content":[{"type":"tool_use","nam e":"Read","input":{"file_path":"/pkg/stor e/store.go"}}],"usage":{"input_tokens":1 5234,"output_tokens":89}}} {"type":"tool_result","content":[{"type": "text","text":"// store.go\npackage store \n\nimport (\n\t\"database/sql\"\n\t\"fmt\"\n )..."}]} {"type":"assistant","message":{"content": [{"type":"text","text":"I see the pattern. // 2.3MB more of this... 454 messages. 2.3MB. Unreadable. ~2ms AFTER (WITH TACIT) Refactor store pattern for archive module Apr 8, 2026 · 47 min · capture/pkg/store · 454 messages INTENT Refactor archive retention to use existing store pattern, not new abstraction layer DECISIONS Chose store.go over new repo Added sync protection before archive delete operations BLOCKERS CGO_ENABLED=1 required for go-sqlite3 (3 failed builds) N+1 query in batch lookup OUTCOMES 23 files modified, 4 packages All tests passing (go test -race) Contract tests added PHASES Research Implementation Debug Harden + Test ARCHITECTURE app_api store archive sqlite BILL OF MATERIALS pkg/store/store.go modified pkg/store/archive.go new desktop/app_archive_api.go modified desktop/contracts/... new + 19 more files... Extraction: ~2ms deterministic, no LLM 454 → 5 sections structured intelligence 1 HTML file self-contained, shareable

HOW IT WORKS

Run agents. Tacit turns the sessions into project memory.

Tacit captures Claude Code, Codex, and Gemini sessions, stores them in a local project index, extracts intent and decisions, then makes the sessions searchable for handoffs and resume context.

LOCAL SESSION PIPELINE

01 / RUN

Start any agent session.

Claude Code, Codex, and Gemini runs appear as local session files.

Claude CodeCodexGemini

02 / CAPTURE

Tacit builds one local index.

Sessions normalize into project-scoped memory without accounts or cloud storage.

project scoped100% localno accounts

03 / EXTRACT

Raw runs become structured memory.

Intent, decisions, blockers, and outcomes become scan-friendly fields.

intentdecisionsblockersoutcomes

04 / REUSE

Search the reason. Resume the work.

Find the exact run, open the handoff, and restart with context.

searchhandoffresume

search("auth decision") -> exact run · handoff · resume context

See what's coming → roadmap

You've run 500 AI-agent sessions.
What did you learn?

“Here's what nobody talks about: Claude Code sessions are incredible in the moment and completely useless 48 hours later. You can't search them, you can't reference them, and you definitely can't remember why you made a particular call three projects ago. I burned real hours re-deriving decisions I'd already made. Tacit is the thing that makes all those sessions compound instead of evaporate. Dead simple. Runs locally. No accounts, no cloud nonsense. Just works.”

Pranav Pranav, Director of Engineering & Founding Member, Velotio Technologies

“I've been on Tacit since the earliest plugin. I ship product across three workstreams, and project intelligence is what makes sense of it — it stitches the decisions I made in one session to the ones I'm making today. Search I use maybe once a week, but when I need it, it pays for itself.”

Honey Arora Honey Arora, Director, Product Management, Rezolve AI

“I love the interface. It's quite convenient and useful. Thanks, take my money :)”

— founder

From the founder

13 research pieces · 7 essays · See all →

PRICING

$29 $49 EARLY ADOPTER

Early adopter price. One-time purchase. Lifetime updates.

  • + Unlimited Claude Code, Codex, and Gemini sessions indexed
  • + Search across all projects
  • + Intent, decisions, blockers captured
  • + Timeline and file change tracking
  • + Export to JSON and Markdown
  • + Windows, Linux, and macOS app
  • + 100% local — no cloud, no accounts
BUY TACIT — $29 →

Secure checkout · No subscription

Not ready to buy?

Try the free version (last 3 days of sessions) →
KK

Ketan Khairnar — staff engineer, 22 years

Built Tacit because I was losing context across hundreds of Claude Code, Codex, and Gemini sessions. Solo. No VC. More →

FAQ

Need help? support@tacit.sh