Skip to content
ContextStreamStart free

Project context for AI agents.

Your agent knows code.Give it the decisions behind yours.

Give your agent the code and context it needs to build on past work. ContextStream connects hosted code search with saved project knowledge—decisions, lessons, and plans your agent can retrieve across sessions through MCP.

Start free

Your favorite agents. On the same page.

Share context, handoffs, and runbooks across the tools you already use.

Save what the code can’t explain.

ContextStream indexes your code for search and keeps your decisions, preferences, insights, and lessons as saved project knowledge. Your agent can retrieve both through MCP, so you don’t have to explain the same context for every task.

Decision

Preserve existing integrations

A choice and its reason: keep the read endpoint because older integrations depend on it.

Ask your agent: “Save this decision: keep the read endpoint because older integrations depend on it.”

Preference

Use the shared UI components

A convention to follow: build new screens with the project’s existing fields and validation.

Ask your agent: “Save this preference: use our shared UI components when building a new screen.”

Insight

Retries can send duplicate notifications

A useful finding: job logs showed the same notification being sent after a retry.

Ask your agent: “Save this insight from the job logs: retries can send duplicate notifications.”

Lesson

Test date filters across time zones

Experience to reuse: check day boundaries in multiple time zones before releasing a date filter.

Ask your agent: “Save this lesson: test date filters across time zones before releasing them.”

See what your agents can do with the whole project behind them.

One project: add project roles without breaking existing integrations.

The implementation agent writes the code. The test agent checks it.

01 / Plan

Plan from the decisions behind the code

TASK
Add project-level roles.
RECORDED DECISION
Check access at the project boundary. Per-route checks duplicated policy logic.
CONSTRAINT
Existing tokens must keep working.
IMPLEMENTATION AGENT“I’ll use the project boundary and test both access paths.”
02 / Build

Carry that direction through the work

PLAN
Keep compatibility in one adapter.
IMPLEMENT
Add project-role checks at the boundary.
TEST
Verify scoped roles and existing-token access.

The saved decision now shapes the code and the tests.

03 / Coordinate

Coordinate when something changes

HUMAN → TEST AGENT“The integration test client needs the old access path until its replacement is ready.”

The test agent records this dependency and sends a coordination notice. The supported implementation agent can retrieve it on its next turn and acknowledge the update.

IMPLEMENTATION AGENT“I’ll add that client’s access case to the tests.”
04 / Save

Make the result useful to others

The implementation agent saves the test outcome, its reason, and the remaining work.

The next agent can retrieve that record instead of reconstructing the session. You can inspect what was saved.

A DECISION IN ONE PART OF THE PROJECT CAN GUIDE WORK ELSEWHERE.

Project context gives agents direction for the work, a way to share what changes, and a place to record what they learn.

Different ways to give your agent project context.

Compare what each approach offers, when it fits, and what you’ll maintain. Choose based on how you work and where your data can be hosted.

ApproachWhat it providesWhat you maintain
Memory and rules in your agent (Agents.MD, Claude.MD)

What it providesInstructions, remembered facts, and decisions available in your agent.Works well when… you mainly use one agent and can maintain its instructions.

What you maintainKeep instructions current and share relevant decisions and outcomes with other agents or people when needed.

Project files, docs, and tickets

What it providesShared plans, requirements, decisions, and work history.Works well when… your team keeps shared records current and can point agents to the relevant ones.

What you maintainKeep records current, point agents to the relevant ones, and record changes and outcomes.

Your own context system

What it providesControl over how context is stored, retrieved, and shared.Works well when… you need custom behavior and can maintain retrieval and integrations.

What you maintainBuild and maintain retrieval, agent integrations, and update delivery.

ContextStream

What it providesCode search, saved project records, and agent saving in one system. Send updates for another supported agent to retrieve on its next turn.Works well when… you want shared code and project knowledge across supported agents, and hosted indexing fits your requirements.

What you maintainConnect your project and agents. Ask agents to save important decisions, keep records current, and review indexing exclusions.

View benchmarks

Keep saved knowledge open to review.

Follow a decision back to its source, revise it, and share the change with agents that have access.

Trace a decision to its source.

Read why invitation links last seven days, alongside the requirements that led to the saved decision.

Update a decision as requirements change.

Ask your agent to save the new 24-hour expiry and its reason. Then check that the saved decision reflects the change.

Share the update with another agent.

Send the revised decision to a connected agent with project access. It can retrieve the update on its next turn; ask it to use the decision when checking the code.

Keep knowledge in the right project.

Choose which project the agent searches. Its workspace and project permissions determine which saved knowledge it can retrieve.

Setup and privacy questions, answered.

How do I connect my agent?

Run setup from your project folder, sign in, choose a workspace, link the project, and select your supported editor. Setup configures the connection and instruction files, then starts indexing in the background. Restart your editor afterward, then ask which project you’re in. Follow the setup guide.

Where does the first context come from?

Your indexed project files provide the initial code context. Saved decisions, lessons, plans, and documents add the reasoning behind the work. Ask your agent to record important knowledge explicitly—indexing cannot recover every undocumented decision.

Is this another documentation job?

Your connected agent can save decisions, lessons, and plans while you work. Ask it to record important outcomes and update them when things change. Capture depends on your workflow; not every conversation becomes durable knowledge automatically.

Can I start on my own?

Yes. Start with one project and one supported agent. Your project knowledge can follow you across sessions and supported editors. Invite authorized collaborators when you need shared access.

Does my code leave my machine?

Yes. Code search sends eligible file contents to ContextStream for hosted indexing, even with a local MCP connection. Review Git ignore rules and .contextignore before indexing. Built-in exclusions help, but they do not guarantee that every sensitive file is excluded.

Who can see the context?

Access depends on the record’s scope, workspace and project permissions, and any explicit sharing. Personal records are not automatically team-visible, and connected services retain their own account permissions. Review membership and external shares before adding sensitive material.

Can I remove stored context?

Yes, but stored records, transcripts, and indexed source have separate removal controls. Turning transcript capture off stops future capture; it does not delete existing history. Review saved decisions and other records separately, and remember that removal cannot retract copies already shared elsewhere.

Try it on a task that matters.

Choose a small task. See whether a saved decision changes how your agent approaches it.

  1. Save one decision

    Connect your project and agent, then ask it to save a decision and its reason.

    Example decision

    Export timestamps in UTC so reports agree across time zones.

  2. Start a fresh session

    Ask the agent to retrieve the project context and add a CSV export. Check whether its plan uses the saved timestamp decision.

    Try asking

    Retrieve the project context, then add a CSV export.

  3. Check the result

    Did the agent apply the decision without another explanation? Ask it to save what the test confirmed.

    What to check
    • The retrieved decision
    • UTC timestamps in the export

One project. One agent.
Your next task.

Start with a decision worth carrying forward.

Start free