Problems get solved twice
Someone fixed the same bug with Claude last month. The fix, and the reasoning behind it, sits on their machine.
NewAsk Cortex: chat with your Chats
Every Claude Code conversation in your company. Saved and searchable.
String Cortex collects the Claude Code sessions from every employee laptop into one secure place, so your team can find how a problem was solved, reuse prompts that worked and see where AI time and cost go.
- Installed for everyone by your Claude admin
- Never slows Claude down
- Only your company’s Claude accounts
0things each employee installs: Claude Code does it
1token per person, pasted once into Claude Code
2ways to search: by keyword and by meaning
3logins: StringHub, company admin, employee
The problem
Your team’s best AI work disappears into laptops
Claude Code writes every session to a file on the laptop where it ran. Nobody else can see it, nobody can search it, and it’s gone when the laptop is.
No view of AI-assisted work
Leads can’t review how code was produced, and compliance can’t answer “what did we send to the AI?”.
Usage is a black box
Which teams, projects and models use the most tokens? Without the transcripts you’re guessing.
New · Ask Cortex
Chat with your Chats
Ask your team’s Claude Code history the way you’d ask the colleague who fixed it. Ask Cortex reads the sessions that match and writes one answer, with numbered sources back to the exact messages. Then keep going: follow-up questions keep the context.
- Answers from your own sessions, every claim linked to its source
- Follow-ups understand “that fix” or “his test” from the chat so far
- Point it at people and projects: @ a colleague, # a project
- Reads only what the person asking may see, nothing more
- Your model, your key: Claude, OpenAI or Gemini, with cost per answer
- How did we fix the Redis lock timeout in #payments-api?
- Why did we drop the v1 refunds endpoint, and what replaced it?
- What have @Priya K. and @Ali M. tried for the flaky checkout tests?
- Find the prompt that wrote our webhook retry tests, I want to reuse it.
- Is migration ka rollback plan kya tha? Hinglish works too.
Features
One searchable home for every session
Everything people type to Claude Code and every reply it gives, organised by project and person, ready to search the moment the reply finishes.
Search by meaning
Type in plain words. Hybrid search mixes keyword matching with semantic search on embeddings made on the server itself, with no outside AI call.
Read whole sessions
Every session as a clean chat, grouped by project. Open up to three side by side.
A work board of sessions
Every session as a card: Kanban by status, person, project or tag, a calendar, or a table. Status follows activity by itself.
Home and profiles
A home page for every role, and a profile for every person with their sessions, projects and a year of activity.
Usage and cost stats
Tokens and API-equivalent cost per day, project, person and model, with model prices you can edit.
Connect once, a workspace per folder
Each person pastes their own token into Claude Code once, and picks a workspace for each new project folder. Make it compulsory if you like.
Organisation lock
Only conversations from your own Claude organisation are accepted. Anything else is refused and listed for review.
Live and reliable
New messages appear within seconds. If the server is down, laptops keep them and catch up on their own.
How it works
Live across your company in an afternoon
No agents to package, no scripts on laptops. Tick your options, your Claude admin pastes the settings once, and Claude Code does the rest.
Sign up on StringHub
Create your organisation, turn String Cortex on and add your Claude organisation id.
Paste the settings
Tick what you want under Claude Code settings and paste the result into your Claude managed settings.
People connect
Each person pastes their own token when Claude Code asks, and picks a workspace for each new folder.
Search everything
From then on every reply is saved, including what waited on the laptop.
Built for teams
The right view for every person
Three kinds of login, each seeing exactly what it should. What each role may do is configurable.
Company admin
e.g. Acme’s engineering lead- All of the company’s conversations
- Sets up workspaces and Claude Code settings
- Decides who sees which projects
- Sees stats for the whole company
Employee
everyone using Claude Code- Uses Claude Code as before, after connecting once
- Logs in to search their own history
- Sees colleagues’ projects only through access groups
- Nothing to install: Claude Code does it
StringHub
runs the service for you- Creates and switches companies on or off
- Watches service health and queues
- Helps with setup and connections
- Never sees any company’s conversations
Security
Built to keep your conversations yours
AI conversations contain your code, your architecture and sometimes your secrets. String Cortex is designed around that.
- HTTPS only; tokens stored as hashes and shown once, and never sent to Claude
- Strict per-company isolation, enforced on every request
- Common secret shapes masked before anything is stored
- Read-only on laptops: transcripts are never changed or deleted
- Switch a company off and its laptops stop at once
| Platform | Organisation | Organisation id |
|---|---|---|
| Claude | Acme | xxxxxxxx-xxxx-… |
| Claude | Acme Labs | zzzzzzzz-zzzz-… |
Refused recently
A conversation from another Claude organisation (yyyyyyyy-yyyy-…) was refused and not kept.
Add this organisation
FAQ
Quick answers
Is this an Anthropic product?
No. String Cortex is built and run by StringHub. It works with Claude Code through Claude Code’s own plugin and settings system.
Do employees need to install anything?
No. On Claude for Teams or Enterprise, your admin pastes the settings String Cortex builds for you into the organisation’s managed settings, and Claude Code installs the plugin on every laptop by itself. Each person then pastes their own token once, when Claude Code asks.
Does it slow Claude Code down?
No. Sending happens in a separate background process after Claude has finished replying, and only the new lines are sent, compressed.
Can employees see each other’s conversations?
Only what an admin shares. An employee always sees their own; access groups can add all projects, chosen projects, or projects with a tag. Company admins see their company’s, and StringHub staff never see any company’s conversations.