Your agent knows your code.
Not the decisions that shaped it.
EmCogni captures engineering decisions at the moment they're made and surfaces them to your agent automatically.
The agent is not wrong.
It just doesn't know what you learned, over time, the hard way.
You ask your agent to speed up a slow endpoint.
“I'll add an index on user_id — that query is doing a full table scan.”
You tried that eight months ago. The index made writes 40% slower on that table. You removed it deliberately. None of that is in the code.
You ask your agent to simplify the payment flow.
“I'll add automatic retries on failed charges — standard reliability pattern.”
Retries caused duplicate charges in production last year. You spent a week on that incident. You added an idempotency layer specifically to avoid this. The agent has no idea.
You ask your agent to refactor two services that share logic.
“I'll move the shared logic into a common module both services can import.”
You deliberately kept them separate after the November outage. Shared state between these two was the root cause. The decision to decouple them is nowhere in the codebase.
Setup flow.
Install and initialize.
One install command, then run init inside your repo. EmCogni writes a pre-commit hook — nothing else to configure.
$ emcogni init
✔ pre-commit hook installed
The hook surfaces context before each commit.
Before every commit emcogni shows what it knows about the staged files and asks if you want to add something. emcogni note works the same way any time between commits.
→ [decision] JWT has no server-side revocation
Your agent won't remember why you made this change.
Capture it? (y/N) y
⬡ Statement: Redis required — JWT can't be revoked server-side
✔ [invariant] stored
Your agent reads it automatically.
Add EmCogni as an MCP server. Your agent calls get_context when it opens a file and propose_context to queue new entries for your review.
"command": "emcogni", "args": ["mcp-run"]
// tools your agent gets
get_context(path) → returns captured context
propose_context(...) → queues a suggestion for review
Same agent. Different experience.
WITHOUT EMCOGNI
WITH EMCOGNI
The agent didn't get smarter. It got context.
Every decision, constraint, and dead end you already explored — your agent now knows. Before you type a word.
Free to start. Pay when you need more.
Local CLI is free forever. All prices exclude tax.
Pro
Your context follows you. Across machines, across wipes, across time.
- ✔Everything in Free
- ✔Unlimited context entries
- ✔Managed cloud DB — no file to babysit
- ✔Works across all your machines automatically
- ✔Full history + git traceability
- ✔Context restore after machine wipes
- ✔Priority support
Free
Free forever. Your machine, your rules.
- ✔Full CLI — all commands
- ✔50 context entries per repo
- ✔Local SQLite store on your machine
- ✔MCP stdio server (Cursor, Claude Code, Copilot)
- ✔You own and manage the file
- ✔Nothing ever leaves your machine
Pro
Your context follows you. Across machines, across wipes, across time.
- ✔Everything in Free
- ✔Unlimited context entries
- ✔Managed cloud DB — no file to babysit
- ✔Works across all your machines automatically
- ✔Full history + git traceability
- ✔Context restore after machine wipes
- ✔Priority support
Team
Your team's decisions, available to every engineer and every agent.
- ✔Everything in Pro
- ✔Unlimited context entries
- ✔Shared context namespace per repo
- ✔All contributors read and propose context
- ✔Owner-approved context proposals
- ✔Shared audit log — who captured what, at which commit
- ✔Repo-level access control
Maintainers of open source projects: apply for a complimentary license →