Atlas / MCP servers / logan-suu / Agora

AgoraBLOCK

mcp/logan-suu/agora

The digital town square where AI agents (PM/Architect/Coder/Tester) collaborate via group chat, while the human acts as the sole Leader. Built with TypeScript + DeepSeek Harness, featuring true parallel execution, cooperative preemption, MCP tools, and git worktree isolation.

Verdict
BLOCK
Grade
F
Trust score
51 /100
Exposed tools
18 12r · 6w · 0d
Transport
stdio
License
—
Stars
3
01

Overview

From the repository's own README, as read at the audited commit. Badges and raw HTML are left out.

A human-led, group-chat workspace where AI agents plan, code, test, and review together.

Contents

  • Current product
  • Product demo
  • What is Agora?
  • Evidence and limits
  • Architecture
  • Quick Start (macOS)
  • Tech Stack
  • Repository Layout
  • Roadmap
  • Design Documents
  • Project Status and Scope

Current scope: local use. Run Agora on your own computer and open its local address in your browser. The backend, Docker sandboxes, and Git worktrees run on that machine, and application data is stored locally. Cloud hosting is outside the current product scope. Model requests go to your chosen online API or local model service; local operation does not mean offline operation. The product installation and startup commands currently support macOS. The Linux product launcher and automatic credential setup are outside that task’s current scope; existing Linux sandbox support is separate.

Planned releases. Phase 10 is complete. Phases 11–13 will introduce Electron, a managed toolchain, and native Mac development, retiring Agora's Docker backend without legacy task migration. Phases 11 and 12 are limited previews; Phase 13 accepts the complete native workflow. These features are planned, not available in the current source build. The instructions below apply to the existing Docker-based implementation. See the roadmap for the smaller release milestones.

Credential setup on macOS. The local launcher creates a random encryption key in your macOS Keychain on first use and reuses it after restart. In the model settings dialog, enter your service's Base URL, API key, and model name, either for one Agent or the whole team. Keychain access may require macOS authorization. Existing encrypted connections remain unavailable when t

Read from source at commit 95c60c23749fOBSERVED · 2026-10-11
02

Connect

Built from this server's own package name, version and transport as found in its source — not copied from anyone's documentation, so it cannot drift against a page we do not control. Replace the environment placeholders with a token scoped to the least it needs.

claude-code
claude mcp add benchmark --env AGORA_CREDENTIALS_KEY=${AGORA_CREDENTIALS_KEY} --env AGORA_FIXED_FAKE_SECRET=${AGORA_FIXED_FAKE_SECRET} --env AGORA_GO_FORMAL_AUTHORIZATION=${AGORA_GO_FORMAL_AUTHORIZATION} --env AGORA_VALIDATION_KEYCHAIN=${AGORA_VALIDATION_KEYCHAIN} -- npx -y [email protected]
claude-desktop
{
  "mcpServers": {
    "benchmark": {
      "command": "npx",
      "args": [
        "-y",
        "[email protected]"
      ],
      "env": {
        "AGORA_CREDENTIALS_KEY": "${AGORA_CREDENTIALS_KEY}",
        "AGORA_FIXED_FAKE_SECRET": "${AGORA_FIXED_FAKE_SECRET}",
        "AGORA_GO_FORMAL_AUTHORIZATION": "${AGORA_GO_FORMAL_AUTHORIZATION}",
        "AGORA_VALIDATION_KEYCHAIN": "${AGORA_VALIDATION_KEYCHAIN}"
      }
    }
  }
}
03

Exposed tools (18)

12 read · 6 write · 0 destructive.

ToolRiskDescription
applyPatchread
checkread
createWorktreeread
diffread
fs_readreadRead task evidence.
inspectreadinspect
listread
mergewrite
probereadReturn a local diagnostic result
readread
read_auditreadRead audit
read_probereadRead the fixture file.
runwrite
sandbox_runwritescripted trace fixture
workspace_applywrite
workspace_readread
workspace_runwrite
writewrite
04

Trust audit

BLOCKgrade F · trust 51/100 Do not install this without reading the findings. The audit found something that could harm you or your machine.

LayerWhat it checksResult
L0Provenance & inventoryPASS
L1Static analysis of the codeFAIL
L2Instruction surface (what it tells the agent)PASS
L3Class-specific surfacePASS
L4Behavioural (sandbox)SKIPPED

What the source does

Filesystem
declared (8 observation(s))
Network
declared (6 observation(s))
Shell
declared (1 observation(s))
Dependencies
not all pinned
Secrets in source
found

Findings (25)

HIGHCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
packages/runtime/sandbox/src/local-command-policy.ts:69
...executables.map((path) => `(allow process-exec (literal ${quote(path)}))`),
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
MEDIUMNetwork egress · net.raw_ip · CWE-200, CWE-319
apps/desktop/src/preview-server.ts:20
const origin = () => `http://127.0.0.1:${(server.address() as AddressInfo).port}`;
MEDIUMHard-coded secrets · secret.generic · CWE-798, CWE-321
apps/web/test/chat-ui.test.ts:193
payload: { secret: 'RAW_PAYLOAD_MUST_NOT_RENDER' },
MEDIUMHard-coded secrets · secret.generic · CWE-798, CWE-321
tests/evals/phase10/local-launcher.eval.ts:170
apiKey: 'startup-test-private-key',
MEDIUMHard-coded secrets · secret.generic · CWE-798, CWE-321
tests/integration/phase10/phase10-model-settings.test.ts:34
const secret = 'test-model-private-key';
MEDIUMHard-coded secrets · secret.generic · CWE-798, CWE-321
tests/integration/phase5/phase5-exit.test.ts:119
payload: { intent: 'forged', secret: 'never-trust-client-payload' },
LOWInventory / provenance · inv.binary · CWE-1104
docs/demo/Agora-Natural-Chat-Artifact-2026-09-12.zip
Agora-Natural-Chat-Artifact-2026-09-12.zip
Why it matters. a compiled or binary member cannot be reviewed from source
Fix. ship source, or explain the binary in the README
LOWInventory / provenance · inv.binary · CWE-1104
docs/reviews/task115-evidence/cleanup/0-output-resource-manifest.json.gz
0-output-resource-manifest.json.gz
Why it matters. a compiled or binary member cannot be reviewed from source
Fix. ship source, or explain the binary in the README
LOWInventory / provenance · inv.binary · CWE-1104
docs/reviews/task115-evidence/cleanup/0-output-service-manifest.json.gz
0-output-service-manifest.json.gz
Why it matters. a compiled or binary member cannot be reviewed from source
Fix. ship source, or explain the binary in the README
LOWInventory / provenance · inv.binary · CWE-1104
docs/reviews/task115-evidence/commit-cleanup/0-output-resource-manifest.json.gz
0-output-resource-manifest.json.gz
Why it matters. a compiled or binary member cannot be reviewed from source
Fix. ship source, or explain the binary in the README
LOWInventory / provenance · inv.binary · CWE-1104
docs/reviews/task115-evidence/commit-cleanup/0-output-service-manifest.json.gz
0-output-service-manifest.json.gz
Why it matters. a compiled or binary member cannot be reviewed from source
Fix. ship source, or explain the binary in the README
LOWInventory / provenance · inv.no_license · CWE-1104
Why it matters. no LICENSE file and no repo licence
Fix. add a licence
LOWInventory / provenance · inv.suspicious_name · CWE-1104
docs/reviews/task123-state-evidence/atomic-invalid-payload-agora-task123-validation-q8aKKH.json
atomic-invalid-payload-agora-task123-validation-q8aKKH.json
Why it matters. member named after an attack tool
Fix. remove or justify
LOWCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
docs/reviews/task112-packaging-evidence/reproduction/historical/main.cjs:15
report.node=exec(res+'/node/bin/node',['--version']);report.pnpm=exec(res+'/node/bin/node',[res+'/pnpm/bin/pnpm.cjs','--version']);report.git=exec(res+'/git/bin/git',['--version']);
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
LOWCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
docs/reviews/task112-packaging-evidence/reproduction/historical/main.cjs:18
exec(res+'/git/bin/git',['init','-b','main',repo]);exec(res+'/git/bin/git',['-C',repo,'config','user.name','Spike']);exec(res+'/git/bin/git',['-C',repo,'config','user.email','[email protected]']);
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
LOWCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
docs/reviews/task112-packaging-evidence/reproduction/main.cjs:16
report.node=exec(res+'/node/bin/node',['--version']);report.pnpm=exec(res+'/node/bin/node',[res+'/pnpm/bin/pnpm.cjs','--version']);report.git=exec(res+'/git/bin/git',['--version']);
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
LOWCode injection · code.eval_exec · CWE-78, CWE-94, CWE-95
docs/reviews/task112-packaging-evidence/reproduction/main.cjs:19
exec(res+'/git/bin/git',['init','-b','main',repo]);exec(res+'/git/bin/git',['-C',repo,'config','user.name','Spike']);exec(res+'/git/bin/git',['-C',repo,'config','user.email','[email protected]']);
Why it matters. evaluates text as code
Fix. remove; use a parser or a dispatch table
LOWFilesystem / path · fs.traversal · CWE-22, CWE-59
apps/desktop/scripts/build.mjs:9
const repo = resolve(dirname(fileURLToPath(import.meta.url)), '../../..');
LOWFilesystem / path · fs.traversal · CWE-22, CWE-59
apps/desktop/src/service.ts:6
import { controlPath, keychainStore } from '../../web/scripts/local-process.mjs';
LOWFilesystem / path · fs.traversal · CWE-22, CWE-59
apps/web/scripts/local.mjs:10
const repo = fileURLToPath(new URL('../../../', import.meta.url));
LOWFilesystem / path · fs.traversal · CWE-22, CWE-59
apps/web/src/app/api/channels/route.ts:1
import { createGetChannels } from '../../../server/message-handlers';
LOWFilesystem / path · fs.traversal · CWE-22, CWE-59
apps/web/src/app/api/channels/route.ts:2
import { messageRuntime } from '../../../server/message-runtime';
LOWNetwork egress · net.env_exfil · CWE-200, CWE-319
tests/integration/phase12/phase12-6.test.ts:109
'.env' ... curl
Why it matters. reads secrets in the same file that sends data out
LOWNetwork egress · net.metadata_ip · CWE-200, CWE-319
packages/runtime/sandbox/test/local-download-policy.test.ts:18
'169.254.169.254',
Why it matters. cloud metadata endpoint: the classic SSRF credential grab
LOWNetwork egress · net.metadata_ip · CWE-200, CWE-319
packages/runtime/sandbox/test/local-download-policy.test.ts:74
new URL('https://metadata.google.internal/', 'https://registry.npmjs.org').href,
Why it matters. cloud metadata endpoint: the classic SSRF credential grab

Gates applied: no_behavioural_pass, no_license.

Audited 2026-10-11 · audit v0.4.1 · source sha 95c60c23749ffull audit observations/trust-audit/mcp-server/logan-suu__agora.json · Report an issue / request a re-scan
05

Audit history

Every audit this server has had. A grade with a past is a grade somebody is still checking.

DateSourceVerdictGradeScoreChange
2026-10-1195c60c23749fBLOCKF51first audit
06

Questions

What is the Agora MCP server?

The digital town square where AI agents (PM/Architect/Coder/Tester) collaborate via group chat, while the human acts as the sole Leader. Built with TypeScript + DeepSeek Harness, featuring true parallel execution, cooperative preemption, MCP tools, and git worktree isolation.

What tools does Agora expose?

18 in total: 12 read-only, 6 that write, and 0 that can delete or overwrite. Every one is listed on this page with its risk.

Is Agora safe to connect to an agent?

No — not without reading the findings first. The audit graded it F (51/100) and found 1 critical or high issue in the source. Each one is listed on this page with the file and line it is on.

What credentials does Agora need?

It reads AGORA_CREDENTIALS_KEY, AGORA_FIXED_FAKE_SECRET, AGORA_GO_FORMAL_AUTHORIZATION, AGORA_VALIDATION_KEYCHAIN and DEEPSEEK_API_KEY from the environment. Give it a token scoped to the least it needs — an agent that can be talked into calling a tool can be talked into calling it with your credentials.

How does Agora run?

It speaks stdio, so it runs as a local process your client starts. It is published on npm as benchmark at 1.0.0.

How current is this page?

The grade is for one exact copy of the source (95c60c23749f), read on 2026-10-11. The repository is watched and re-audited when it changes.

Advertisement