Modem

Context engineering

Engineer the context your agents run on

Stop hand-feeding your agents. Modem is the memory and context layer that keeps them current on what customers want and what you're building.

  • Superwhisper
  • Kernel
  • Sentry
  • Zumper
  • OpenCode
  • Polytomic
  • Fig
  • Composio

Your customers have thoughts.
Most of them get lost.

  • Finding it is hard

    Support sees the email, sales hears it on a call, product reads it in Slack. Nobody notices it's the 10th time this month.

  • Fixing it is slow

    Engineering finally gets a ticket, with half the context never captured. Scoping and fixing it both take longer.

  • Following up rarely happens

    The fix ships. Whoever heard the complaint never finds out, so the customer who asked never hears back.

Every concern, grouped and linked to the PRs that fix it

Modem groups what customers say into topics, then tracks the issues and pull requests working on each one. You always know what’s open, in progress, or completed.

The practice

Curate the context. Don't cram it.

everything pasted inthe merged topic, sized to fit

Curated beats crammed

Pasting every thread into a prompt burns the context window before the agent starts. Modem hands over the merged topic instead, sized to fit.

session contextcleared every rundurable customer memorytopicspeoplecompaniesquotes

Two kinds of memory

Session context clears when the run ends. Customer memory should not. Modem keeps topics, people, and companies durable, so every new session starts current.

a plausible guessagent without context12 reports · 3 channels“the 25k exporttimes out” · Soniaagent with the graph

The same question, grounded

Ask an agent what customers think of exports and it guesses. Give it the graph and it answers with counts, affected accounts, and the exact quotes.

What it is

What context engineering means for your agents

An agent is only as good as what it can see. Context engineering is the work of deciding what that is, for every run. Modem does the customer half of that work. It reads Slack, Discord, support tickets, sales calls, and issues, merges them into a customer context graph, and serves the merged picture to any agent that asks over the Modem MCP server. Your agents stay current without anyone pasting threads into prompts.

  • Current, not snapshotted

    The graph updates as feedback arrives, so an agent asking today gets today, not the state of the export from last quarter.

  • Shared across agents

    Claude Code, Cursor, and the Modem agent in Slack all read the same graph, so every tool answers from the same facts.

  • Pull, not paste

    Agents query for what a task needs over MCP instead of receiving everything up front, which keeps the context window for the work.

FAQ

Common questions about context engineering

What is context engineering?
Context engineering is deciding what an agent sees before it starts work. Instead of pasting raw threads into a prompt, you give the agent curated, current context. For customer-facing work that means the merged picture of what customers said, who they are, and what they pay for.
How is agent memory different from RAG?
RAG retrieves documents that look similar to the question. Memory holds structured facts that stay true between sessions, like which customers hit a bug and what plan they are on. Modem maintains that structure for you, so a query returns joined facts instead of lookalike text.
How do agents connect to Modem?
Over MCP. Run `claude mcp add --transport http modem https://mcp.modem.dev/mcp` in Claude Code, or register the same endpoint in Cursor, Codex, or any MCP-compatible agent. Claude, ChatGPT, and Cursor also have an official Modem plugin. For scripted access there is an official CLI on npm, @modem-dev/cli.

Less digging, more shipping

Research takes minutes, and every fix comes with the list of who asked. Your agents self-serve answers faster through Modem's MCP instead of reading raw threads.

  • 3–4 hrs to 10 min

    Product research tasks that used to take an afternoon1

  • 5–10 times

    Customer touchpoints surfaced per bug fix, across every channel1

  • 98% fewer tokens

    Agents read Modem's context graph vs the raw threads underneath2

  1. Research time and touchpoints per fix: reported by Sentry's product team. Read the Sentry case study
  2. 98% fewer tokens: measured by Modem on real usage data, on average.

Trusted by teams that listen (and ship)

  • Kernel
  • Superwhisper
  • OpenCode
  • VLT
  • Sentry
  • Zumper
  • Mycroft
  • Fig
  • Novelcrafter
  • HumanLayer
  • Basedash
  • Polytomic
  • Composio
  • FusionAuth

“Things that would have taken three or four hours are ten-minute tasks now, and when we fix something, Modem shows us everyone who was waiting.”

Nicholas MeisenheimerProduct Manager @ Sentry
Read Case Study

“While I’m working on something I can just hop over and ask Modem, and get an answer in minutes – instead of searching across all the channels and getting out of the flow.”

Leonie GräßelFounder @ Novelcrafter
Read Case Study

Connects to every channel and tool, answers wherever you work

Modem reads the tools your customers and team already use, then hands the whole story to people and agents, sources attached. Explore all integrations

Chat

  • Slack
  • Discord
  • Teams
  • Email

Support

  • Intercom
  • Zendesk
  • Front
  • Plain
  • Pylon
  • Jira Service Desk

Calls

  • Gong
  • Fathom
  • ZoomSoon

Customers

  • Salesforce
  • Stripe
  • Canny

Analytics

  • PostHog
  • Mixpanel
  • Amplitude
  • Sentry

Code

  • GitHub
  • GitLab
  • Linear
  • Jira

Need an integration we don't have? Talk to us

  • Dashboard

    Navigate your customer graph using a complete web UI.

  • Agent

    Full-featured cloud agent, available in Slack, Teams, Discord, or web.

  • Plugins & MCP

    Bring your favorite desktop, terminal, or cloud agent.

  • CLI

    Query your Modem data from any terminal. Great for agents.

  • HTTP API

    The same data as everything else, as JSON over HTTP.

Security and compliance

Modem combines industry-standard compliance, enterprise-grade controls, and privacy-first product features at every layer.

  • SOC 2

    SOC 2 compliant. Independently audited against the AICPA Trust Services Criteria.

    Trust center →
  • SSO + SCIM

    Manage access at scale with SSO and SCIM provisioning from the identity provider you already run.

  • Governance

    Audited agent runs, roles, PII scrubbing, and custom retention.