aleph

Aleph is a Python MCP server and CLI/skill that implements a Recursive Language Model (RLM) workflow: it persists working state (loaded contexts, search indexes, evidence, sessions) outside the LLM prompt window, supports iterative search/navigation, server-side code execution over loaded context (exec_python), and recursive sub-query orchestration to converge on answers.

Evaluated Mar 30, 2026 (21d ago)
Homepage ↗ Repo ↗ Ai Ml mcp agent rlm recursive-reasoning tooling python self-hosted
⚙ Agent Friendliness
56
/ 100
Can an agent use this?
🔒 Security
49
/ 100
Is it safe for agents?
⚡ Reliability
38
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
70
Documentation
70
Error Messages
0
Auth Simplicity
65
Rate Limits
20

🔒 Security

TLS Enforcement
80
Auth Strength
35
Scope Granularity
25
Dep. Hygiene
55
Secret Handling
55

Self-hosted MCP server suggests transport security depends on client/server setup (TLS not described in README). Auth/authorization for the MCP interface is not documented; provider/API keys are configured via environment variables for sub-query backends. Given exec_python capability, the primary risk is code execution/sandboxing and data exposure; README mentions caps/truncation and an ALEPH_CONTEXT_POLICY=isolated option for more defensive defaults.

⚡ Reliability

Uptime/SLA
10
Version Stability
55
Breaking Changes
50
Error Recovery
35
AF Security Reliability

Best When

You want an agent workflow that loads large context once, iterates with MCP tools to narrow scope, runs computation server-side, and persists results across long sessions.

Avoid When

You cannot control the safety of code execution/output exposure, or you need a fully managed hosted service with strong reliability guarantees.

Use Cases

  • Large log or document analysis with iterative search and evidence collection
  • Codebase navigation and investigation (search/peek/chunk, then compute/verify)
  • Data exploration over loaded context (run Python over files/data, return derived results)
  • Long-running investigations that need session persistence and resume
  • Recursive problem solving by splitting tasks into sub-queries and recipes

Not For

  • Untrusted/hostile workloads where server-side code execution (exec_python) cannot be sandboxed
  • Environments needing a simple stateless API-only integration without local process/session management
  • Use cases requiring guaranteed availability with published SLAs (it appears primarily local/self-hosted)

Interface

REST API
No
GraphQL
No
gRPC
No
MCP Server
Yes
SDK
No
Webhooks
No

Authentication

Methods: Environment-variable configuration for provider backends (e.g., ALEPH_SUB_QUERY_API_KEY for OpenAI-compatible sub-query API; ALEPH_LLAMACPP_URL for local llama.cpp) No explicit auth documented for the MCP server in provided README
OAuth: No Scopes: No

The README describes authentication for certain sub-query backends via environment variables (e.g., ALEPH_SUB_QUERY_API_KEY) but does not describe MCP-server authentication/authorization controls.

Pricing

Free tier: No
Requires CC: No

Open-source (MIT) package; likely self-hosted. Provider costs depend on the selected backend (OpenAI-compatible, Claude CLI, or local llama.cpp).

Agent Metadata

Pagination
none
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • exec_python runs code over loaded context; ensure sandboxing and strict input/output handling when used by autonomous agents
  • get_variable('ctx') is described as policy-aware but retrieving full raw context can defeat the intended context-minimization strategy
  • Recursive sub-query backends depend on the selected install profile; misconfiguration of provider/model env vars can break recursion

Alternatives

Full Evaluation Report

Comprehensive deep-dive: security analysis, reliability audit, agent experience review, cost modeling, competitive positioning, and improvement roadmap for aleph.

AI-powered analysis · PDF + markdown · Delivered within 30 minutes

$99

Package Brief

Quick verdict, integration guide, cost projections, gotchas with workarounds, and alternatives comparison.

Delivered within 10 minutes

$3

Score Monitoring

Get alerted when this package's AF, security, or reliability scores change significantly. Stay ahead of regressions.

Continuous monitoring

$3/mo

Scores are editorial opinions as of 2026-03-30.

8642
Packages Evaluated
17761
Need Evaluation
586
Need Re-evaluation
Community Powered