cloud context infrastructure

Your files,
any AI agent

Your AI agents can't see your files. ctxd connects your cloud data and makes it searchable via API, CLI, or MCP.

from ctxd import Client

client = Client()
results = client.search(
    "Q4 revenue projections",
    apps=["google-drive", "slack"]
)

for r in results:
    print(f"{r.title} ({r.app}, {r.modified})")

$ pip install ctxd
$ ctxd login
$ ctxd connect google-drive
✓ Connected. Indexing 1,847 files...

$ ctxd search "deployment runbook" --app github
Found 3 results:
  1. deployment-runbook.md (github/backend, 2d ago)
  2. Deploy process PR #142 (github/infra, 1w ago)
  3. Runbook discussion (slack/#engineering, 3d ago)

{
  "mcpServers": {
    "ctxd": {
      "command": "ctxd",
      "args": ["mcp"],
      "env": { "CTXD_API_KEY": "your-api-key" }
    }
  }
}

Works with Claude, Cursor, or any MCP-compatible agent.

Trusted by engineers at

how it works

Your apps

OAuth connects in minutes

ctxd_

Parse. Index. Search. Sync.

PDFs, docs, code, images (OCR)

Any AI agent

API, CLI, or MCP

Claude ChatGPT Cursor Your agent

four ways to use ctxd

mcp / for agents

Any MCP-compatible agent calls ctxd natively.

Your agent

ctx

MCP server

Your apps

claude_desktop_config.json

"ctxd": { "command": "ctxd", "args": ["mcp"] }

cli / for developers

Search from the terminal.

$ ctxd search "deployment runbook" --app github

Found 3 results:

deployment-runbook.md

github/backend · 2d ago

0.94

Deploy process PR #142

github/infra · 1w ago

0.87

Runbook discussion

slack/#engineering · 3d ago

0.82

api / for your code

REST API for search and retrieval.

Request

client.search(
  "Q4 revenue projections",
  apps=["google-drive", "slack"]
)

200 OK · 12 results · 48ms

Q4 Revenue Model.xlsx

google-drive · Oct 12

0.96

#finance: Q4 projections thread

slack · Oct 10

0.91

skills / for automation

Composable actions defined in SKILLS.md.

# weekly-digest Active

Slack, Drive, GitHub · past 7 days

Summarize

Group by team, highlight blockers

Post to #team-updates

Every Monday 9am

All four hit the same index. Same data, same permissions.

why ctxd

Not another assistant. The context layer for any AI agent.

01 / not another assistant

Every agent upgraded

You already use Claude, ChatGPT, Cursor. ctxd makes all of them better.

Claude ChatGPT Cursor → ctxd_

02 / human to LLM format

UI content to searchable markdown

PDFs, Slack threads, Google Docs. ctxd converts UI-structured content into clean text.

"user": "U04QAH..."
"ts": "171103..."
"blocks": [...]

→

Sarah (10:15)
deploy blocked
#engineering

03 / search that compiles

One query. Zero noise.

Private data has no crowd signals. ctxd's DSL filters by app, date, channel, folder.

"runbook" AND app:github
AND date:>2026-03-01

04 / all connectors included

Budget goes to agents, not plumbing

Other tools charge per connector. ctxd includes everything.

$0 extra

05 / minutes, not months

Ship the agent, not the pipeline

OAuth + parsing + indexing takes months. ctxd replaces it in an afternoon.

$ ctxd connect google-drive
✓ Indexing 1,847 files...
✓ Ready.

06 / always current

Agents get the latest version

Files change. Index syncs. No cron jobs. No stale answers.

→ runbook.md updated 2m ago

→ index synced 1m ago

→ agent gets latest now

connectors

Google Drive

Docs, Sheets, Slides, PDFs, images. Full folder hierarchy preserved.

Slack

Messages, threads, file attachments across channels.

GitHub

Repositories, issues, pull requests, code files.

Google Calendar

Events and scheduling data for time-aware queries.

More coming. Request an integration →

security

Google CASA Tier 2 certified

→ TLS 1.2+ in transit, encrypted at rest

→ Per-user data isolation

→ OAuth2 scoped tokens

→ User data deletable on request

→ On-prem deployment available

→ SOC 2 Type II on roadmap

from the blog

March 17, 2026

What is ctxd?

ctxd stands for context daemon. A background process that turns your cloud files into searchable context for AI agents.

March 12, 2026

What humans see vs what LLMs see

Your files are formatted for humans. LLMs need clean text. The gap is why AI agents hallucinate.

March 7, 2026

How ctxd powers enterprise agentic workflows

Multi-step agents need context from every app. Here is the architecture.

All posts →

pricing

Free to start. All connectors included.

Pro from $15/user/mo. No per-connector fees. Ever.

See pricing details →

you might be wondering

We can build this ourselves. +

You can. OAuth, parsing, chunking, indexing, sync, permissions. That is months of infrastructure before your agent makes its first query. ctxd connects in minutes.

We already have internal search. +

Your internal search has a UI. Your agents can't call it. Your CI pipeline can't query it. ctxd exposes search via API, CLI, and MCP. Anything that writes code can use it.

Is this another AI assistant? +

No. ctxd is the infrastructure underneath. It gives context to any assistant you already use. Claude, ChatGPT, Cursor, your own agent.

What about hallucination? +

LLMs answer correctly less than 43% of the time on fact-seeking questions. Every ctxd result links to its source. Your agent gets grounded context, not guesses.

What about data security? +

CASA Tier 2 certified. Per-user isolation. OAuth-scoped. Encrypted in transit and at rest. Deletable on request.

What if a file changes or gets deleted? +

ctxd tracks changes in source apps continuously. The index stays current. No cron jobs. No stale data.

How is this different from other context or memory tools? +

Most context tools store conversations. ctxd indexes your actual files. Most memory tools need you to push data in. ctxd pulls from your apps directly via OAuth. And most tools lock search inside their own UI. ctxd exposes it via API, CLI, and MCP.

Get started for free

Connect your apps. Index everything. Query from anywhere.

Get started

Your files,any AI agent