sift-gateway

Sift is a reliability gateway that captures JSON outputs from upstream MCP/CLI tools, persists them as SQLite artifacts, redacts secrets on the way back to the model, and returns either inline JSON (`full`) or stable schema-based references (`schema_ref`) for follow-up. It provides explicit pagination continuation until completion and a Python-based artifact querying interface for exact retrieval across one artifact or an entire pagination chain.

Evaluated Mar 30, 2026 (0d ago)
Homepage ↗ Repo ↗ Ai Ml ai-ml agent-tooling mcp cli gateway artifacts json pagination security schema sqlite
⚙ Agent Friendliness
57
/ 100
Can an agent use this?
🔒 Security
54
/ 100
Is it safe for agents?
⚡ Reliability
29
/ 100
Does it work consistently?

Score Breakdown

⚙ Agent Friendliness

MCP Quality
20
Documentation
85
Error Messages
0
Auth Simplicity
60
Rate Limits
10

🔒 Security

TLS Enforcement
30
Auth Strength
55
Scope Granularity
40
Dep. Hygiene
65
Secret Handling
85

The README claims outbound secret redaction is enabled by default and that it keeps full payloads out of prompt context; it also references a security policy and an error contract in docs. TLS enforcement and scope granularity are not evidenced in the provided text. Persistent SQLite artifact storage may retain sensitive data, so configuration and retention controls are important.

⚡ Reliability

Uptime/SLA
0
Version Stability
45
Breaking Changes
30
Error Recovery
40
AF Security Reliability

Best When

You have MCP/CLI tools that return large or paginated JSON and you want agents to do exact, schema-stable follow-up queries without leaking secrets into model context.

Avoid When

You cannot store tool outputs locally (SQLite) or cannot comply with data retention policies for those artifacts; also avoid if you need a simple stateless request/response API.

Use Cases

  • MCP tool output reliability for LLM agents (stable schemas, pagination-complete retrieval)
  • Handling large/JSON-heavy tool results without context stuffing by storing full payloads as artifacts
  • Secret-safe outbound redaction before returning data to the model
  • Repeatable analysis over paginated data using code executed against stored artifacts
  • Agent workflows needing provenance/reproducibility of tool outputs across multi-step runs

Not For

  • Use as a general-purpose data warehouse or long-term analytics platform (it appears primarily an artifact store for agent/tool runs)
  • Environments where persistent local SQLite artifacts are disallowed (no evidence of managed remote storage)
  • When you need a network REST/GraphQL API surface instead of a CLI/MCP integration

Interface

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

Authentication

Methods: Upstream login (token storage backend) Authentication tokens configured via deployment/config (details in docs/SECURITY)
OAuth: No Scopes: No

The README/manifest mention Google ADC and an OAuth-token storage backend used by `upstream login`, but the external-surface auth mechanism is not fully specified in the provided text (no explicit API-key/OAuth flow details for the gateway itself).

Pricing

Free tier: No
Requires CC: No

Open-source package (MIT) distributed via PyPI; no pricing information provided.

Agent Metadata

Pagination
explicit-multi-page-contract
Idempotent
False
Retry Guidance
Not documented

Known Gotchas

  • Agents must continue pagination explicitly until `pagination.retrieval_status == COMPLETE`; stopping early can yield partial answers.
  • Be careful to query the correct root path (e.g., `$` vs `metadata.usage.root_path`) to avoid analyzing only a subset of nested results.
  • Ensure secret redaction/allow-lists align with your policy; storing artifacts increases data retention risk if misconfigured.

Alternatives

Full Evaluation Report

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

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.

6533
Packages Evaluated
19870
Need Evaluation
586
Need Re-evaluation
Community Powered