MCP ExtensionIncluded with Pro & PAYG

Your IDE already has one AI. MegaLens makes it sharper.

MegaLens MCP plugs into Claude Code, Codex, Cursor, and other MCP editors to add planning, research, code intelligence, security audit, architecture review, blindspot detection, gap-finding, and final closure review without replacing your existing workflow.

MegaLens is the specialist layer. Your host IDE stays the final judge.

Can save you thousands in review costs

MegaLens can catch code, security, and architecture issues early, before they turn into expensive fixes or outside audit bills.

Cuts wasted output and token spend

MegaLens trims bulky MCP output so your IDE gets the useful findings without paying for as much extra noise.

Puts less load on your IDE

Tested with up to 35% estimated overall context savings when MegaLens handles specialist work before handing it back to your IDE.

Try it with just $0.50

Verify your card for $0.50 and get $5.50 total credit, enough to properly test planning, audits, and MCP workflows.

How it works

Three steps to sharper IDE workflows.

01

Connect MegaLens to your IDE

Add one MCP config block. Claude Code, Codex, Cursor, and similar MCP clients can connect in seconds.

02

Send targeted specialist work

Use MegaLens for planning, audits, architecture review, research, blindspot detection, and gap-finding instead of stuffing all of that into your main IDE thread.

03

Let the host IDE make the final call

MegaLens returns tighter specialist findings. Your host IDE decides what to accept, change, reject, or act on next.

Workflow fit

MegaLens complements the host IDE.

The point is not another heavy interface. The point is to move specialist work into a separate MCP layer so your main IDE assistant has less context pressure and fewer wasted cycles.

Claude Code

Already in flow

Claude is already in the host seat

MegaLens adds

Specialist work from outside the host model family

Result

MegaLens adds planning, audit, blindspot checks, and Slice B compaction while Claude Code stays the final judge.

G

Codex (OpenAI)

Already in flow

Codex is already driving the session

MegaLens adds

Targeted specialist review without replacing Codex

Result

MegaLens handles parallel specialist work so Codex can keep more reasoning room for final action and closure.

C

Cursor / Other MCP

Already in flow

Your editor already has a primary assistant

MegaLens adds

A specialist layer for planning, audit, research, and gap-finding

Result

MegaLens complements the host assistant instead of becoming another IDE. The host model remains the final decision-maker.

Behind the scenes

What MegaLens MCP is doing in the background.

In simple terms: MegaLens does the specialist thinking first, sends back tighter structure, and leaves the host IDE with less waste and a cleaner final review job.

1. MegaLens takes the specialist work out of the main IDE thread

Problem: Your IDE gets overloaded when it has to plan, audit, structure, and execute everything in one thread.

What MegaLens does: MegaLens moves that specialist work into a separate MCP layer first.

Benefit: Your IDE carries less load, so it stays sharper and wastes fewer tokens on setup thinking.

2. Different specialists look at the problem from different angles

Problem: One model can miss risks simply because it sees the problem from one angle only.

What MegaLens does: Planner and review engines are used for gap-finding and blindspot reduction. A code specialist like Devstral focuses on turning findings into a practical execution plan.

Benefit: Blind spots are reduced earlier, which can save money on missed code, security, and architecture issues later.

3. The planner is pushed to fill gaps, not create noise

Problem: Long AI answers often look impressive but hide the useful part inside noise.

What MegaLens does: MegaLens pushes the planner toward missing risks, missing steps, and missing edge cases instead of rewarding long verbose output.

Benefit: You pay for fewer wasted output tokens, and your IDE gets more signal with less fluff.

4. Findings come back structured instead of bloated

Problem: A giant wall of text is expensive to pass back into the IDE and hard to act on.

What MegaLens does: MegaLens shapes the result into findings, priorities, and execution steps.

Benefit: Structured output lowers input token waste because your IDE can work from clean steps instead of re-reading everything.

5. Your IDE does the final review, not the heavy lifting

Problem: If the IDE has to re-do the whole audit itself, you lose time and token budget.

What MegaLens does: MegaLens handles the specialist planning and structure first, then hands a tighter result back to the host IDE.

Benefit: The IDE can focus on final judgment and gap fill instead of doing all the heavy lifting from scratch.

6. Step-by-step plans reduce execution waste too

Problem: Execution gets expensive when the IDE keeps stopping to figure out the next move.

What MegaLens does: Devstral and the other specialist layers return a clearer step-by-step execution plan.

Benefit: The IDE spends less effort deciding what to do next, which can lower execution cost as well as review cost.

7. Internal compaction keeps context cleaner over time

Problem: As conversations grow, context gets bloated and the IDE becomes less efficient.

What MegaLens does: MegaLens compacts conversation and MCP payloads internally so context stays tighter over time.

Benefit: This improves context hygiene over time. It is one contributor to the overall efficiency gain, not the whole reason behind the up to 35% estimated savings.

Setup

Two lanes, one engine

Local MCP for Pro / BYOK users who want the most reliable IDE workflow. Hosted HTTP for PAYG users who want the quickest setup with managed billing.

Local MCP

Pro / BYOK

Your key, your machine

Hosted HTTP

PAYG / Evaluation

Managed billing, no install

0

Pick your lane

a.

Pro / BYOK — bring your own OpenRouter key, run local MCP on your machine. Best for daily reliability and deeper IDE workflows.

b.

PAYG — verify your card for $0.50, get $5.50 total credit ($0.50 deposit + $5.00 bonus), connect via hosted HTTP. Quickest way to evaluate MegaLens.

Pro / BYOK
1

Local MCP

Runs on your machine via stdio transport with your own OpenRouter key. No dependency on hosted round-trip latency or client timeout behavior — the most reliable setup for daily IDE use.

Claude Code

Add to ~/.claude.json or your project's .mcp.json:

{
  "mcpServers": {
    "megalens": {
      "type": "stdio",
      "command": "npx",
      "args": ["-y", "@megalens/mcp@latest"],
      "env": {
        "MEGALENS_OPENROUTER_KEY": "sk-or-v1-your-key"
      }
    }
  }
}
Codex (OpenAI)

Add to your project's codex.json or workspace config:

{
  "mcp": {
    "servers": {
      "megalens": {
        "type": "stdio",
        "command": "npx",
        "args": ["-y", "@megalens/mcp@latest"],
        "env": {
          "MEGALENS_OPENROUTER_KEY": "sk-or-v1-your-key"
        }
      }
    }
  }
}
2

Hosted HTTP MCP

PAYG

Quickest setup — no local install, managed billing via MegaLens credits. Connect to the hosted endpoint with your token and start evaluating immediately.

Grab your MegaLens token from Settings → MCP Server:

ml_tok_••••••••••••••••

Or use the CLI helper to auto-detect your editor and write the hosted config:

npm install -g megalens-mcp
megalens-mcp setup
Claude Code

Add to ~/.claude.json or your project's .mcp.json:

{
  "mcpServers": {
    "megalens": {
      "url": "https://megalens.ai/api/mcp",
      "headers": {
        "Authorization": "Bearer ml_tok_your_token_here"
      }
    }
  }
}
Codex (OpenAI)

Add to your project's codex.json or workspace config:

{
  "mcp": {
    "servers": {
      "megalens": {
        "url": "https://megalens.ai/api/mcp",
        "headers": {
          "Authorization": "Bearer ml_tok_your_token_here"
        }
      }
    }
  }
}
3

Start asking for targeted work

MegaLens tools appear in your IDE automatically. Ask for specific specialist work:

> Use megalens to audit the auth middleware for security issues
> Use megalens to review this PR for architecture and blind spots
> Use megalens to compact this MCP output with Slice B before handing it back

Core scopes

Fast to scan. Clear in seconds.

Planning, research, code intelligence, security audit, architecture review, blindspot detection, gap-finding, and closure review.

Planning

Use MegaLens for serious planning passes before commits, migrations, infra changes, or product decisions.

Research

Targeted research can happen in a separate MCP layer instead of bloating the host IDE thread.

Code intelligence

Repo-aware specialist review sees real files, imports, dependencies, and structure instead of a pasted summary.

Security audit

Useful for catching issues earlier, when they are cheaper to fix than after release or after a late-stage audit.

Architecture review

Good for tradeoff checking, blindspot coverage, and gap-finding when one model feels too certain too early.

Final closure review

Use MegaLens at the end to check what still feels unresolved before the host IDE makes the final call.

Without MegaLens vs MCP

The point is not more noise. The point is better use of context.

FeatureWithoutWith MegaLens MCP
Uses your existing IDE workflow
Host IDE remains final judge
Blindspot coverage from a specialist layer
Slice B compaction for verbose output
Planning and audit support
Repo-aware code intelligence
Context efficiency focus
PAYG entry point

Sharper workflow. Less waste.

Under a few dollars for meaningful planning or audit work can be negligible next to shipping mistakes. External audits can cost $2000+.